JP2001014307A - Document processing apparatus, document processing method, and recording medium - Google Patents

Abstract

(57) [Summary] [Problem] To realize reading out by a plurality of characters and display according to the reading out. SOLUTION: For example, a plurality of types of voices respectively corresponding to a plurality of persons are selectively used to generate a voice signal serving as a reading voice for electronic document data, and according to a type of the generated voice signal. To output a speaker image signal. Thereby, for example, the speaker can be switched in the middle of the document data of the explanatory sentence, and the presentation in the conversation state is realized in the document data as the conversation sentence.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to electronic document processing, and more particularly to a document processing apparatus capable of outputting a reading voice of an electronic document.

[0002]

2. Description of the Related Art Conventionally, WWW (World Wide Web) has been provided as an application service for providing hypertext information in a window format on the Internet.

[0003] The WWW is a system for executing document processing for creating, publishing, or sharing a document, and showing the way of a new style document. However, from the viewpoint of practical use of documents, advanced document processing beyond WWW, such as classification and summarization of documents based on the contents of the documents, is required. For such advanced document processing, mechanical processing of the contents of the document is indispensable.

However, mechanical processing of the contents of a document is still difficult for the following reasons. First
HTML, a language that describes hypertext
(Hyper Text Markup Language) specifies the expression of a document, but hardly specifies the content of the document. Second, a network of hypertexts formed between documents is not always easy for a reader of the document to understand the contents of the document. Third
In general, the author of a document writes without considering the convenience of the reader, but the convenience of the reader of the document is not adjusted to the convenience of the author.

[0005] As described above, the WWW is a system showing the way of a new document. However, since the document is not mechanically processed, advanced document processing cannot be performed.
In other words, in order to perform advanced document processing, it is necessary to process the document mechanically.

Therefore, a system for supporting mechanical processing of documents has been developed based on the results of natural language research, with the goal of mechanical processing of documents. As a document processing based on natural language research, mechanical document processing using tags attached to a document has been proposed on the assumption that attribute information about the internal structure of the document by the author of the document, so-called tags are added. I have.

[0007]

By the way, with the recent spread of computers and the progress of networking, text processing, creation of text documents, labeling, change, and the like have been performed by using an index depending on the contents of the documents. There is a demand for enhanced document processing to be performed. For example, it is desired to summarize a document according to a user's request, classify the document, read out the document, and the like. In addition, processing more suitable for the user, such as display styles suitable for the form of the document viewer or display unit that displays documents and their summaries, read-aloud functions performed with display, and interlocking of display of documents and images with read-aloud. It has been demanded.

[0008]

SUMMARY OF THE INVENTION The present invention has been proposed in view of the above-described circumstances, and it is an object of the present invention to provide a document processing apparatus that reads out a document by effectively using a display output and a voice output. Aim.

For this reason, the document processing apparatus according to the present invention comprises: a read-out voice generating means capable of selectively using a plurality of types of voices to generate a voice signal serving as a read-out voice for electronic document data; Image generating means for outputting a specific type of speaker image signal according to the type of the audio signal generated by the means. In particular, the apparatus further includes a display unit, a voice output unit, and a control unit. The control unit causes the voice output unit to output the voice signal generated by the read-out voice generation unit, and generates the voice signal generated by the read-out voice generation unit. Control is performed so that the display operation on the display means is performed using the speaker image signal output from the image generation means according to the type of the sound signal being output. That is, according to the present invention, in addition to reading out an electronic document, a character image or the like as the speaker can be displayed, and a plurality of types of voices and speaker images are output, so that the reading out more suitable for the contents of the electronic document can be performed. Make the action possible.

[0010] The reading voice generating means selects the type of the generated voice signal based on the voice type information added to the electronic document data. Further, the image generating means is a still image signal, a moving image signal, or a pseudo moving image as a speaker image signal corresponding to the type during a period when a certain type of audio signal is output from the reading voice generating means. Output a signal.

According to the document processing method of the present invention, there is provided a reading voice generating procedure for generating a voice signal as a reading voice by selectively using a plurality of types of voices for electronic document data, and outputting the voice signal from a voice output unit; An image generating procedure for outputting a speaker image signal corresponding to the type of the audio signal generated by the voice generating procedure and displaying the signal on the display unit is performed. Further, it is assumed that the recording medium of the present invention has recorded therein an operation control program having such procedures.

Further, the document processing apparatus of the present invention selects one voice from a plurality of types of voices based on voice type information added to the electronic document data, and uses the selected voice to process the electronic document data. A read-out voice generating means capable of generating a voice signal serving as a read-out voice of
Image generating means for selecting one speaker image from a plurality of types of speaker images based on the voice type information and outputting an image signal based on the selected speaker image. Further, according to the document processing method of the present invention, one voice is selected from a plurality of types of voices based on voice type information added to the electronic document data, and the read-out voice of the electronic document data is selected using the selected voice. A speech signal generating procedure for generating a speech signal to be output from the speech output unit, and selecting one speaker image from a plurality of types of speaker images based on the speech type information, and an image based on the selected speaker image And an image generating procedure for outputting a signal and displaying the signal on the display unit. Further, it is assumed that the recording medium of the present invention has recorded therein an operation control program having such procedures.

[0013]

Embodiments of the present invention will be described below in the following order. 1. 1. Configuration of document processing device 2. Document data structure 3. Manual classification process for document data 3-1 Processing procedure 3-2 Index creation 3-3 Document browsing / classification creation / classification operation 3-4 Classification model creation / registration 4. Automatic classification processing for document data 4-1 Processing procedure 4-2 Automatic classification Summary creation processing Speech processing

1. 1. Configuration of Document Processing Apparatus As shown in FIG. 1, a document processing apparatus 1 according to an embodiment of the present invention includes a main body 10 having a control unit 11 and an interface 12, and a main body 10 which receives an input from a user.
, An input unit 20 for sending and receiving signals to and from the outside, and a display unit 30 for displaying an output from the main body 10
Recording / reproduction for recording / reproducing information on / from the recording medium 32
It has a reproduction unit 31, an audio output unit 33, and an HDD (hard disk drive).

The main body 10 has a control unit 11, an interface 12, and an image memory 16, and constitutes a main part of the document processing apparatus 1. The control unit 11 has a CPU 13 that executes processing in the document processing apparatus 1, a RAM 14 that is a volatile memory, and a ROM 15 that is a non-volatile memory. For example, the CPU 13
According to the procedure recorded in the OM 15, if necessary, the data is temporarily stored in the RAM 14 and control for executing the program is performed. The operation of the control unit 11 will be described in detail later, but includes a classification process for the supplied document data, a summary creation process, a process of generating a speech reading file for a reading operation,
And document analysis required for these processes. The programs and application software necessary for these operations are stored in the ROM 15, the HDD 34, or the recording medium 32. The document processing program used by the control unit 11 is stored in a ROM
15 or a recording medium 32 or HDD
34, for example, the communication unit 2
It is also conceivable to download a document processing program provided by an external server or the like from a network such as the Internet via the Internet.

The image memory 16 is a part for storing various images such as a human image in order to display a speaker image on the display unit 30 during a reading operation described later. The control unit 11 can configure a window image using the respective image data stored in the image memory 16 and display the window image on the display unit 30. It should be noted that the image memory 16 is
1 internal memory. Alternatively, the image memory 1
6 is not provided, and the function of the image memory 16 is
M15 may be provided.

The interface 12 includes a control unit 11, an input unit 20, a communication unit 21, a display unit 30, and a recording / reproducing unit 3.
1. Connected to the audio output unit 33 and HDD. Under the control of the control unit 11, the interface 12 inputs data from the input unit 20, inputs and outputs data to and from the communication unit 21, outputs data to the display unit 30, and controls the recording / reproducing unit 31. Each operation of data input / output, data output to the audio output unit 33, and data input / output to the HDD 34 is performed. More specifically, it adjusts the timing of data input / output between the control unit 11 and each of the above units, converts the format of data, and the like.

The input section 20 is a section for receiving an input from the user to the document processing apparatus 1. This input unit 20
Is composed of, for example, a keyboard and a mouse. The user uses the input unit 20 to input characters such as keywords using a keyboard, or to display the display unit 30 using a mouse.
For example, the user can select an element of the electronic document that is displayed in. Hereinafter, an electronic document handled by the document processing apparatus 1 is referred to as “document data” or simply “document”. The “element” is an element constituting a document, and includes, for example, a document, a sentence, and a word.

The communication unit 21 is a unit that receives a signal transmitted from the outside to the document processing apparatus 1 via, for example, a communication line or transmits a signal to the communication line. The communication unit 21 receives one or a plurality of data such as documents transmitted from the outside, and sends the received data to the main body 10. Of course, data is transmitted to an external device via a communication line.

The display section 30 is a section for displaying characters and image information as an output of the document processing apparatus 1. The display unit 30 is, for example, a cathode ray tube (CR).
T), a liquid crystal display (LCD), or the like, for example, displays one or more windows, and displays characters, figures, and the like on these windows.

The recording / reproducing section 31 records / reproduces data on / from a recording medium 32 such as a so-called floppy disk. Here, the recording medium 32
An example of a floppy disk (magnetic disk) is given as an example of an optical disk, a magneto-optical disk,
Recording medium 3 if it is a portable medium such as a memory card.
2 can be applied as an example. Then, the recording / reproducing unit 31
Any recording / reproducing device (disk drive, card drive, etc.) according to the media may be used.

When the recording medium 32 has recorded therein a document processing program for processing a document,
The recording / reproducing unit 31 can read out the document processing program from the recording medium 32 and supply it to the control unit 11. If document data is recorded on the recording medium 32, the recording / reproducing unit 31 can read it out and supply it to the control unit 11. That is, for the document processing apparatus 1,
This is an input mode of the document data different from the reception of the document data by the communication unit 21. Further, the control unit 11 can cause the recording / reproducing unit 31 to record the document data processed by the document processing apparatus 1 on the recording medium 32.

The voice output unit 33 is a part for outputting a document as an output of the document processing apparatus 1 as a reading voice. That is, when a voice signal generated by a voice synthesis process based on document information (a reading file to be described later) is supplied by the control unit 11, the voice output unit 33 performs output processing of the voice signal to perform display. Together with the unit 30, it functions as an output unit of the document processing apparatus 1.

The HDD 34 provides a large-capacity recording area in the document processing apparatus 1. The HDD 34 includes the control unit 11
The recording / reproduction of information is performed based on the control of. This HDD
Reference numeral 34 denotes an application program for various processes executed by the control unit 11, for example, a program for speech synthesis, and the like, and stores, for example, document data and the like taken into the document processing apparatus 1. It can be used as a part to be kept.

2. Document Data Structure Next, the structure of document data in this example will be described. In this example, document processing is performed with reference to a tag that is attribute information assigned to the document. The tags used in this example include syntactic
c) There are tags and semantic / pragmatic tags that enable mechanical understanding of the content of documents between multiple languages.

Some syntactic tags describe the internal structure of a document. As shown in FIG. 2, the internal structure by tagging is configured such that each element such as a document, a sentence, and a vocabulary element is linked to each other by a normal link and a reference / referenced link. In the figure, a white circle “○” indicates an element, and the lowest white circle is a vocabulary element corresponding to the minimum level word in the document. The solid line is a normal link indicating a connection between elements such as a document, a sentence, and a vocabulary element, and the dashed line is a reference link indicating a dependency relationship by reference / reference. The internal structure of a document is document (documemt), subdivision (subdivision), paragraph (paragraph), sentence (sentenc
e), subsentential segment (subsentential s
egment), ..., vocabulary elements. Of these, subdivisions and paragraphs are optional.

On the other hand, as the semantic / pragmatic tagging, there is a method of describing information such as a meaning like a meaning of a polysemy. Tagging in this example is performed by HTML (Hyper Te
XML (Extensible Mar) similar to xt Markup Language
kup Language).

An example of tagging is shown below, but tagging a document is not limited to this method. In the following, examples of English and Japanese documents are shown, but the description of the internal structure by tagging can be similarly applied to other languages.

For example, "Time flies like an arro
The sentence "w." can be tagged as follows. <> is a tag added to the document.

<Sentence><noun phrase meaning = “time0”> time </ noun phrase><verbphrase><verb meaning = “fly1”> flies <// verb><adverbphrase><adverb meaning = like0> like < / Adverb><
Noun phrase> an <noun meaning = “arrow0”> arrow </ noun></ noun phrase></ adverb phrase></ verb phrase>. </ Sentence>

Here, <sentence>, <noun>, <noun phrase>, <
Verbs, <verb phrases>, <adverbs>, <adverb phrases> are sentences, nouns, noun phrases, verbs, verb phrases, adjectives / adverbs (including prepositional or postpositional phrases), adjective phrases / adverbial phrases, respectively. Is shown. In other words, the syntactic structur of a sentence
e).

These tags are arranged immediately before the front end and immediately after the end of the element. The tag placed immediately after the end of the element indicates that it is the end of the element by the symbol "/". Elements are syntactic constructs: phrases, clauses, and sentences. Note that the meaning (word sense) = “time0” corresponds to the word “ti”.
It refers to the plural meanings of “me”, that is, the 0th meaning of the plural meanings.
Has at least the meanings of a noun, adjective, and verb. Here, it indicates that the word "time" is a noun (= the 0th meaning). Similarly, the word "orange" has at least the meaning of plant name, color, and fruit, but these can also be distinguished by their meaning.

In this example, the syntactic structure of the document data can be displayed in the window 101 on the display unit 30 as shown in FIG. In this window 101, vocabulary elements are displayed on the right half 103, and the internal structure of the sentence is displayed on the left half 102.

For example, as shown in FIG.
1 has a sentence “A” in which the internal structure is described by tagging.
In C City, where his B meeting was over, some popular and general newspapers have stated on paper that they will voluntarily regulate their photographic coverage. Is displayed. An example of tagging this document is as follows:

<Document><sentence><adverb phrase relation = “place”><nounphrase>
<Adverb Phrase Location = "C City"><Adverb Phrase Relation = "Subject"><Noun Phrase Identifier = "B
Kai ”><adverb phrase relation =“ affiliation ”><person name identifier =
"Mr. A"> Mr. A </ personal name></ adverb phrase><organization name identifier = "B association"> B association </ organization name></ noun phrase></ adverb phrase> finished </ adverb Phrase><place name identifier = “C city”> C city </ place name></ noun phrase>, </ adverb phrase><adverb phrase relation = “subject”><noun phrase identifier = “press” syntactic =
“Parallel”><nounphrase><adverbphrase> Some </ adverb phrase> popular paper </ noun phrase> and <noun> general paper </ noun></ noun phrase></ adverb phrase><adverb Phrase relation = “object”><adverb phrase relation = “contents” subject = “press”><adverb phrase relation = “object”
><Nounphrase><adverbphrase><noun co-reference = "B-kai"></ adverb phrase of </ noun></ photo ad </ noun phrase>
Adverb phrase> Self-regulated </ adverb phrase> Policy </ adverb phrase><adverb phrase interrelation = "position"></ adverb phrase> on paper. </ Text></text>

By being tagged in this way, the structure of the document is represented by each pair of tags <> to </>. For example, <document> to </ document> indicates the range of one document, and similarly, <sentence> to </ sentence> indicates the range of one sentence. Also, for example, <Noun phrase identifier = “B meeting”>
According to </ noun phrase>, the part "Mr. A's B meeting"
It is expressed as a noun phrase with "kai" as the identifier. That is, the internal structure of the sentence shown on the left half surface 102 in FIG. 3 is expressed by the tagging.

Further, in this document, "part of popular paper and general paper" is expressed in parallel by the tag "syntax =" parallel ". The definition of parallel is to share a dependency relationship. If nothing is specified, for example, <noun phrase relation = x><noun> A </ noun><noun> B </ noun></ noun phrase> indicates that A has a dependency on B Represent. Relation = x represents a relation attribute.

The relation attribute describes an interaction between syntactic, meaning, and rhetorical. Grammar functions such as subjects, objects, and indirect objects, subject roles such as an actor, a subject, a beneficiary, etc., and rhetorical relations such as a reason and a result are described by the relation attributes. In this example, the subject, object,
Describe relative attributes for relatively easy grammar functions such as indirect objects.

In this document, "Mr. A",
For proper nouns such as "B Association" and "C City", place names,
Attributes are described by tags such as person names and organization names. By adding tags such as place name, person name, organization name, etc.,
It is expressed that the word is a proper noun.

3. Manual Classification Processing on Document Data 3-1 Processing Procedure In the document processing apparatus 1 of this example, when document data is taken in from the outside by, for example, the communication unit 21 (or the recording / reproducing unit 31), the document data is converted according to the content. Perform classification processing. In the following description, it is described that document data from the outside is taken in via the communication unit 21.
The description is the same when the document data is supplied from the outside in the form of a portable medium such as a floppy disk and the document data is taken in from the recording / reproducing unit 31.

As the classification process, there are a manual classification process in which the user manually classifies according to the contents of the document data, and an automatic classification process in which the document processing apparatus 1 automatically classifies. These classification processes are performed based on a classification model described later. However, in the document processing apparatus 1, no classification model exists in an initial state. Therefore, at the time of the initial state, a classification process including creation of a classification model is required as a manual classification process. After the classification model is generated, the input document data can be automatically classified. First, a manual classification process that needs to be performed first will be described. That is, the manual classification process means that when the document processing apparatus 1 in the initial state receives document data sent from the outside, the control unit 11 creates a classification model and classifies the document data based on a user operation. Is performed.

FIG. 4 shows the overall processing procedure as the manual classification processing. The detailed processing of each processing step will be described later.

Step F11 in FIG.
2 shows a document receiving process by the receiving unit 21 of FIG. In step F11, the receiving unit 21 receives one or a plurality of documents transmitted via, for example, a communication line. The receiving unit 21 sends the received document to the main body 10 of the document processing device. The control unit 11 stores the supplied document data or document data in the RAM 14 or the HDD 34.

In step F12, the control section 11 of the document processing apparatus 1 extracts the features of the plurality of documents sent from the receiving section 21 and creates feature information, that is, an index of each document. The control unit 11 stores the created index in, for example, the RAM 14 or the HDD 34.
As will be described later, the index includes proper nouns, meanings other than proper nouns, and the like, which are characteristic of the document, and can be used for document classification and search.

The document browsing in step F13 is a process executed as required by the user. That is, it is performed according to a user operation. Step F13 and the next step F14 are processes based on a user operation.
The user can view the content of the input document data on the screen of the display unit 30 by performing a required operation. While browsing the document, the user can instruct various processes such as summarization, which will be described later, by operating the icons and the like on the screen. The process will proceed to the classification operation. In step F14, in response to the user performing an operation of setting a classification item (the classification item is also referred to as a category in this specification), the control unit 11
Generates / displays classification items. The user also performs an operation of sorting the document data into the set classification items, and the control unit 11 sorts and displays the document data accordingly.

In step F15, the control unit 11 creates a classification model according to the classification item creation and classification operation performed by the user in step F14. The classification model includes a plurality of classification items (categories) for classifying documents, and associates each document index with each category index (index of each document created in step F12) to change the classification state. This is the specified data.
After generating such a classification model, step F16
Then, the classification model is registered. That is, the control unit 11 performs the registration by storing the classification model in, for example, the RAM 14. By the processing of FIG. 4 described above, manual classification and creation of a classification model have been performed for one or a plurality of pieces of document data input when the document processing state 1 is in the initial state. The processing of step F12 and subsequent steps in FIG. 4 will be described in detail.

3-2 Index Creation In step F14, the control section 11 creates an index for the input document data. First, a specific example of an index created for certain document data will be described.

<Index date = "AAAA / BB / CC"
Time = “DD: EE: FF” Document address = “1234”><User operation history Maximum summary size = “100”><Number of selected elements = “10”> Picturetel </
Selection> ... </ User operation history><Summary> Tax reduction scale, not touching-Prime Minister X's interview </ Summary><Word Meaning = "0003" Central activity value = "140.6"> No touch </ word><Word meaning = “0105” identifier = “X” central activity value =
“67.2”> Prime Minister </ word><person name identifier = “X” word
Meaning = "6103" Central activity value = "150.2"> Prime Minister X /
Word / person name><word meaning = “5301” central activity value = “120.6”> determined </ word><word meaning = “2350” identifier = “X” central activity value =
“31.4”> Prime Minister </ word><word meaning = “9582” Central activity value = “182.3”> emphasized </ word><word meaning = “2595” Central activity value = “93.6”> touch </ word><Word Teijin = “9472” Central activity value = “12.0”> foretold </ word><word Sense = “4934” Central activity value = “46.7”> not touched </ word><word Sense = “0178” Central activity value = "175.7"> Explained </ word><word Meaning = "7248" Identifier = "X" Central activity value =
“130.6”> I </ word><word Meaning = “3684” Identifier = “X” Central activity value =
“121.9”> Prime Minister </ word><word meaning = “1824” Central activity value = “144.4.”> Appealed </ word><word meaning = “7289” Central activity value = “176.8”> showed Word></index>

In this index, <index> and </ index> indicate the start and end of the index, <date> and <time> indicate the date and time when this index was created, <summary> and </ Summary> indicates the start and end of the summary of the contents of this index, respectively. Also, <word> and <
/ Word> indicates the beginning and end of the word. Further, for example, the meaning = “0003” indicates that it is the third meaning. The same applies to other cases. As mentioned above,
Since the same word may have a plurality of meanings, a number is predetermined for each meaning in order to distinguish them, and the corresponding meaning is represented by a number.

<User operation history> and </ User operation history> are the start and end of the user's operation history, and <Select> and </ Select> are the start and end of the translated element. Are respectively shown. The maximum summary size = “100” indicates that the maximum size of the summary is 100 characters, and the number of elements = “10” indicates that the number of selected elements is 10.

As in this example, the index includes proper nouns, meanings other than proper nouns, and the like, which are characteristic of the document. For example, the process of step F12 for creating such an index will be described with reference to FIGS. FIG. 5 shows an index creation process for one document data. Therefore, when a process is performed for a plurality of document data, the process of FIG. 5 is performed for each document data. Further, FIG. 7 shows a detailed process of step F31 in FIG. 5, and FIG. 8 shows a detailed process of step F43 in FIG.

In the index creation processing in step F12 in FIG. 4 described above, first, active diffusion in step F31 in FIG. 5 is performed. This active diffusion is a process in which the central activity value of an element is spread based on the internal structure of a document with respect to document data, so that an element related to an element having a high central activity value also has a high central activity value. It is. That is, after giving a central activity value as an initial value to each element constituting the document, the central activity value is diffused based on the internal structure of the document, specifically, the link structure. Since this central activity value is determined according to the internal structure by tagging, it is used for extracting features of a document. The control unit 11 determines in step F3
As 1, the active diffusion is performed, and the central active value of each element obtained as a result of the active diffusion is stored in, for example, the RAM 14.
Will be stored.

FIG. 6 shows the active diffusion in step F31.
8 will be described in detail. First, an example of a link structure between elements is shown in FIG. In FIG.
The periphery of elements E1 and E2 is shown as part of the structure of the elements and links that make up the document. E1
E8 is an example of an element, in which element E
1 and E2 will be described.

It is assumed that the central activity value of the element E1 is e1, and the central activity value of the element E2 is e2. The elements E1 and E2 are connected by a link L12 (the normal link or the reference link described above). The end point of the link L12 connected to the element E1 is T12, and the end point of the link L12 connected to the element E2 is T21. Element E1 further includes elements E3, E4,
E5 is connected to links L13, L14, and L15, respectively. The end points of the links L13, L14, and L15 on the element E1 side are denoted by T13, T14, and T, respectively.
It is assumed to be 15. Element E2 is composed of elements E6 and E6.
Links L26, L27, L2 for E7 and E8 respectively
8 are connected. Each link L26, L27, L28
At the element E2 side are denoted by T26 and T26, respectively.
27 and T28. The active diffusion processing of FIGS. 7 and 8 will be described using an example of such a link structure.

In step F41 of FIG. 7, when starting the active diffusion for the document data to be indexed, the control unit 11 first initializes the central active value for all elements of the document data. As the initial value of the central activity value, for example, a high value is given to a proper noun, an element selected (clicked) by a user, or the like. Further, the control unit 11 sets the end point activation value of the end point T (xx) of the link connecting the elements to 0 for the reference link and the normal link. The control unit 11 stores, for example, the RAM 14 with the initial value of the endpoint activation value thus assigned.

In step F42, the control unit 11
Initializes a counter for counting the elements Ei constituting the document. That is, the count value i of the counter for counting elements is set to one. If i = 1, this counter will refer to the first element (eg, element E1 in FIG. 6).

In step F43, the control unit 11
Executes a central activation value update process for calculating a new central activation value for the element referenced by the counter. This center activation value update process will be described in detail with reference to FIG. 8 by taking the process for the element E1 as an example. The center activation value updating process is a process of updating the end point activation value of the element, and calculating a new center activation value using the updated end point activation value and the current center activation value.

In step F51 of FIG.
Is an element Ei (for example, in this case E
In 1), a counter for counting the number of links having one end connected thereto is initialized. That is, the count value j of the counter for counting the links is set to one. When j = 1, this counter refers to the first link L (yy) connected to the element Ei. In the example of FIG.
For example, the link L12 is referred to as the first link for the element E1.

In step F52, the control unit 11 determines whether or not the link being referred to, that is, the link L12 connecting the elements E1 and E2, is a normal link by referring to the tag of the related attribute. If the link L12 is a normal link, the control unit 11 proceeds to step F53, and if the link L12 is a reference link, proceeds to step F54.

If the link L12 is determined to be a normal link and the process proceeds to step F53, the control section 11 performs a process of calculating a new endpoint activation value of the endpoint T12 connected to the normal link L12 of the element E1. Endpoint T12
Are the endpoint activation values of all the endpoints connected to the link other than the link L12 among the endpoint activation values of the linked element E2 (in this case, T26 and T2).
7, each end point activation value t26, t27, t28 of T28)
And the central activation value e2 of the element E2, and the value obtained by this addition is divided by the total number of elements included in the document. The control unit 11 performs such an operation using each endpoint activity value and each center activity value read from the RAM 14, thereby calculating a new endpoint activity value for the endpoint connected to the normal link. The calculated end point activation value is stored in the RAM 14. That is, the end point T1
The end point activation value t12 of No. 2 is updated.

On the other hand, when it is determined in step F52 that the link L12 is a reference link and the process proceeds to step F54, the control unit 11 similarly sets a new end point activation of the end point T12 of the element E1 connected to the normal link L12. The processing for calculating the value is performed. The calculation for calculating the end point activity value is as follows. That is, in this case, the endpoint activation values t12 of the endpoint T12 are the endpoint activation values of all the endpoints connected to the link other than the link L12 among the endpoint activation values of the linked element E2 (in this case, T2
6, each end point activation value t26, t27, t of T27, T28
28) and the central activation value e2 of the element E2. (That is, the point where there is no division is different from the case of the above-mentioned normal link.) The control unit 11 performs such an operation using each of the end point activation values and each of the center activation values read from the RAM 14, A new endpoint activity value is calculated for the endpoint connected to the reference link, and the calculated endpoint activity value is stored in the RAM.
14 is stored. That is, the end point activation value t1 of the end point T12
Update 2.

After performing the processing in step F53 or F54, the control unit 11 performs the processing in step F5 through the determination processing in step F55 (if the determination result is NO).
Proceeding to 7, the count value j is incremented, and the process returns to step F52. That is, subsequently, by setting the count value j = 2, the second link (for example, link L13) for the element E1 is referred to.
In the same manner as described above, the end point activation value t13 of the end point T13 connected to the link L13 is calculated / updated in the processing after step F52.

In step F55, the control unit 11 determines whether or not a new endpoint activation value has been calculated for all the links for the element Ei (E1) currently referred to by the count value i, and branches the processing. Because
The process of updating the endpoint activation values is performed until all endpoint activation values of the element Ei being referred to are updated. That is, by repeating the process while incrementing the count value j in step F57, for the element E1, for example, the endpoint activation values t12, t13, t14, and t15 are updated for the endpoints T12, T13, T14, and T15, respectively. When all of them have been updated, the process proceeds from step F55 to F56.

In step F56, a new central activity value ei of the element Ei is calculated by using the updated endpoint activity values in response to the determination of all the endpoint activity values for the element Ei. The new central activation value ei of the element Ei is the current central activation value e of the element Ei.
It is determined by the sum of new endpoint activation values of all endpoints of i and element Ei. For example, in the case of the element E1 in FIG. 6, the new central activation value e1 (new) is e1 (new) = e1 + t12 + t13 + t14 + t15.

The control unit 11 calculates the central activation value ei of the element Ei being referred to by the current count value i in this way. Then, the control unit 11 causes the RAM 14 to store the calculated new central activation value ei. That is, the central activation value ei of the element Ei is updated. (However, at this point, the old center activation value is also held because it is used in the process of step F45 described later.)

When the processing shown in FIG. 8 is performed as the central activation value updating processing in step F43 in FIG. 7, the processing of the control unit 11 proceeds to step F44 in FIG.
The control unit 11 determines whether or not the central activation value update processing has been completed for all elements in the document. Specifically, control unit 11 determines whether or not count value i has reached the total number of elements included in the document. Control unit 1
When the central activation value update processing has not been completed for all the elements, the process proceeds to step F47, where the count value i is incremented, and
Return to For example, after the processing for the element E1 is completed as described above, the count value i is set to 2 and the element E2 is referred to this time. Then, for the element E2, the central activation value update processing of step F43 (that is, the processing of FIG. 8) is performed in the same manner as described above. Although the details will not be described because they are redundant description, in the case of the link example of FIG. 6, in the case of the element E2, the end point activation values t21, t26, t27 of the end points T21, T26, T27, T28 in the processing of FIG. After t28 is updated,
A new central activity value e2 (new) is calculated as e2 (new) = e2 + t21 + t26 + t27 + t28, and is updated.

In the process shown in FIG. 7, the count value i is incremented in step F47, and the central activation value updating process in step F43 is repeated while changing the reference element. Will be updated.

When the updating of the central activation value has been completed for all the elements in the document, the processing is terminated at step F44.
To F45. In step F45, the control unit 11 calculates the average value of the change in the central activity values of all the elements included in the document, that is, the change in the newly calculated central activity value from the original central activity value. For example, the control unit 11 reads the old central activation value stored in the RAM 14 and the updated new central activation value for all elements included in the document. Then, for each element, the difference between the new central activity value and the old central activity value is obtained, and the sum of the differences is divided by the total number of elements to calculate the average value of the change in the central activity value of all elements. The control unit 11 causes the RAM 14 to store, for example, the average value of the change in the central activity value of all elements calculated in this way.

Subsequently, at step F46, the control unit 11
Determines whether the average calculated in step F45 is within a preset threshold. When the average value is within the threshold value, the control unit 11 ends a series of steps as the active diffusion process. When the average value is not within the threshold value, the control unit 11 returns to step F42 and performs the above-described series of processes. Perform the process again.

This series of active diffusion processing can be said to be processing for raising the central activity value of an element related (linked) to an element having a high central activity value. However, by performing this activation diffusion only once, when considering the purpose of the index creation processing, among the elements that should be able to raise the central activation value,
In some cases, the central activity value may not be sufficiently raised. For example, in one activation diffusion, for an element directly linked to an element for which the initial value of the central activation value is set high, the central activation value is raised to some extent, but for an element that is not directly linked, it is Even if the element is important as an index, the central activity value may not be sufficiently raised. Therefore, by performing the active diffusion process a plurality of times as necessary based on the determination in step F46, the central activity value can be converged as a whole, and important elements that cannot be increased in the central activity value As much as possible. The reason why the central activity value is converged as a whole in the active diffusion process a plurality of times is that the central active value of each element updated in the active diffusion process is based on the central activity value of each element, and further the active activity diffusion process is performed for each element. This is because the central activity value is updated. However, if such activation diffusion processing is performed too many times, the central activation values of all the elements converge and become almost the same, which is inappropriate.
For this reason, in the processing of steps F45 and F46, the end timing of the active diffusion process is determined based on the change so as to obtain the average value of the change of the central active value, so that the active diffusion suitable for index creation is performed. Will be realized.

When the active diffusion processing as shown in FIGS. 7 and 8 (ie, step F31 in FIG. 5) is completed, the controller 1
The process 1 proceeds to step F32 in FIG. In step F32, the control unit 11 determines in step F3
Based on the central activity value of each element obtained in step 1, an element whose central activity value exceeds a preset threshold is extracted. The control unit 11 causes the RAM 14 to store the extracted elements.

Subsequently, in step F33, the control section 11 reads out the elements extracted in step F32, for example, from the RAM 14. Then, the control unit 11 extracts all proper nouns from the extracted elements and adds them to the index. Proper nouns have special properties such as having no meaning and not appearing in dictionaries, and are therefore treated separately from words other than proper nouns. As described above, the word meaning corresponds to each of a plurality of meanings of the word. Whether or not each element is a proper noun can be determined based on a tag attached to the document. For example, in the internal structure by tagging shown in FIG. 3, "A", "B" and "C"
“City” is a proper noun since the related attributes of the tags are “person name”, “organization name”, and “place name”, and the control unit 11 adds the extracted proper noun to the index, and The result is stored in the RAM 14.

In the next step F34, the control unit 1
1 extracts the meanings other than proper nouns from the elements extracted in step F32, adds them to the index, and stores the result in the RAM 14.

By the above processing, for example, an index as in the above specific example is generated. That is, the index finds the features of the tagged document and arranges the features, and the features of the document are determined based on the central activity value diffused according to the internal structure of the document. It will be. Since such an index includes meanings and proper nouns representing features representative of a document, it can be used when referring to a desired document. The index includes, in addition to the meanings and proper nouns representing the characteristics of the document, the document in the RAM 14 (or H
The document address indicating the position stored in DD34) is included.

3-3 Document Browsing / Category Creation / Category Operation The index creation processing described above with reference to FIGS. 5 to 8 is performed in step F12 of FIG. Therefore, FIG.
As for the manual classification processing of step F13 and step F1
The process proceeds to the process of No. 4, that is, the process of browsing and manual classification by the user as described above.

As described above, in step F13 of FIG. 4, the user can browse the document displayed on the display unit 30. In step F14, the user can perform an operation of setting a classification item and an operation of sorting document data into the set classification item. The operations performed in steps F13 and F14,
The corresponding process of the control unit 11 and a display example of the display unit 30 are as follows.

FIGS. 9 and 10 show specific examples of the display on the display unit 30. FIG. First, FIG. 9 is a display example of a document classification window 201 corresponding to a classification model described later in detail. That is, a graphic user interface (graphic user interface) used for displaying a document classification.
GUI). This document classification window 201
The operation button display 202 includes a position reset button 202a for returning a screen window state to an initial position, a browser button 202b for calling a browser (browser) for reading the contents of a document, An exit button 202c is displayed.

The document classification window 301
Document classification areas 203, 204, 205,... Are formed as small windows corresponding to the classification items corresponding to the classification model. The document classification area 203 is an area for displaying “other topics”. The “other topics” document classification area 203 is an area where documents that have not been classified yet are presented. For example, step F in FIG.
Each document received at 11 (that is, a document to be classified) is presented in the "other topics" document classification area 203. The document classification area 204 is an area where documents classified as, for example, “business news” are presented. The document classification area 205 is an area in which documents classified as, for example, “political news” are presented. In addition to these, document classification areas without reference numerals in the figure are areas where documents corresponding to specific classification items are presented.

Each of these document classification areas 203 and 204
.., The documents classified into the classification items (categories) set in the respective document classification areas are presented by the icon of the document and the title of the document. If there is no title, a summary of one sentence is displayed. Also, the size of each of the document classification areas 203, 204,... Is not fixed,
When the user moves the delimiting frames 211, 212, 213,... Separating the document classification areas by a drag operation or the like, the document classification areas 203, 204,.
Can be arbitrarily changed. The user can arbitrarily increase or decrease the number of document classification areas.

Each of the document classification areas 203, 204,.
The title (for example, "political news") can be arbitrarily set and changed by the user. Note that the number of document classification areas and each title correspond to the classification items of the classification model described later. In other words, the user sets or deletes a document classification area or performs a title setting in the classification window 201 by operating the input unit 20 using a mouse, a keyboard, or the like.
The number and title of the classification items of the classification model are set.

FIG. 10 shows an example of a browsing window 301 for a user to browse the contents of document data. For example, the user clicks on a certain document in the classification window 201 of FIG. 9 to select the document, and then clicks the browser button 202b, so that the control unit 11
Opens a browsing window 301 for displaying the selected document as shown in FIG.

In the browsing window 301, a file name display section 3 for displaying the file name of the document data file
02, a document display unit 303 for displaying the document data of the file name, a summary display unit 304 for displaying a summary of the document displayed on the document display unit 303, and a keyword display unit 305 for inputting / displaying a keyword. . As the operation button display 306, a summary creation button 306a for instructing creation of a summary sentence, an undo button 306b for performing an undo operation (operation cancellation), a reading button 306c for executing a reading operation, and the like are displayed. Is done.

In such a browsing window 301,
The user can browse the document displayed on the document display unit 303. If the entire document cannot be displayed, a part of the document is displayed. Of course, you can browse the full text by scrolling. In addition, by clicking the summary creation button 306a, the user can cause a summary sentence for the document displayed on the document display unit 303 to be created and displayed on the summary display unit 304. The process of the control unit 11 for creating the summary sentence will be described later. Further, by clicking the read-aloud button 306c, the user can cause the text or the summary of the document displayed on the document display section 303 to be read aloud. This reading operation will also be described later.

The classification window 201 and the browsing window 301 as described above are displayed on the display unit 20 at any time according to the user's operation, not limited to the manual classification processing in FIG. As for the manual classification process,
The user can confirm the type and content of the received document in the classification window 201 and the browsing window 301. Specifically, one or a plurality of documents received in step F11 in FIG. 4 are subjected to the index creation processing in step F12, and then the classification window 2 shown in FIG.
01 "other topics" document classification area 20
3 is displayed. In this classification window 201,
The user manually classifies each of the documents displayed in the document classification area 203. If, for example, the content is not known only by the title of the document, the user checks the content of the document by using the browsing window 301 in FIG. . The browsing performed as needed by the user in this manner is the process of step F13 in FIG.

In step F14, the user can arbitrarily add, update, or delete a classification item on the classification window 201. In response to the operation, the control unit 11 causes the displayed document classification area 203, 204 ...
・ Change the display mode (number, area, title, etc.). The setting / change of the classification item (title of the document classification area) by the user is reflected in the classification model described later.

After setting the classification items as necessary, the user sorts the documents displayed in the document classification area 203 into the respective document classification areas. That is, the documents are classified manually by the user. Specifically, the icon of the document displayed in the document classification area 203 of “other topics” is dragged to the document classification area corresponding to a desired classification item (category) using, for example, the mouse of the input unit 20. Do it by doing. For example, the user sets a document classification area titled “Sports”, and changes the icon of a sports-related document displayed in the document classification area 203 of “Other topics” to the document classification area of “Sports”. Perform an operation like dragging. The icons and titles of the documents that have been manually classified in this manner are thereafter displayed in the dragged destination document classification area.

3-4 Creation / Registration of Classification Model As described above, when the user performs the manual classification operation, the control unit 11 determines in step F15 in FIG. 4 that the classification model is composed of a plurality of categories based on the user's classification operation. Create That is, the control unit 11 collects indices of the plurality of documents classified into each category and generates a classification model. Then, the plurality of documents are classified into each category of the classification model.

The classification model is composed of a plurality of classification items (categories) for classifying documents. Then, for each category, a data format is provided in which the classified documents are shown.
An index is formed for each document in step F12 and the like, and the classification model is, for example, as shown in FIG.
As shown in (1), the data structure is such that the indexes of the documents classified for each category are associated with each other.
In FIG. 11A, the category is “sports”.
“Company”, “Computer”, etc. are set, but these are the classification items set by the user in the classification window 201 as described above. Of course, even if the user does not set, it is set in advance (that is, displayed as a document classification area in the classification window).
There may be categories. Each of the classification items is associated with an index IDX1, IDX2,.
That is, the index of the document classified by the user as described above is associated with each classification item.

The index associated with each classification item is the index of a document displayed in the document classification area of the classification item in the classification window 201. For example, the index IDX1 is associated with the category "sports" because the user creates a document classification area with the title "sports" in the classification window 201, and furthermore, sets the icon of the document of the index IDX1 to "sports". Is manually dragged to a document classification area having a title of "".

As described above, the index of each document includes proper nouns, meanings other than proper nouns, document addresses, and the like. Then, for example, as shown in FIG.
One or a plurality of indices are associated with one classification item. Since the indexes include proper nouns, meanings, document addresses, and the like, the classification model can be represented as shown in FIG. 11B.

That is, as shown in FIG. 11B, the classification model has a structure having columns of proper nouns, meanings other than proper nouns, and document addresses as category indexes corresponding to each category. Then, in the classification model, the proper noun “Mr. A” for each category “sports” “society” “computer” “plant” “art” “event”
... "," Mr. B, ... "," Company C, Company G, ...
"," D type, ... "," Mr. E, ... "and" F
A proper noun such as "Mr." is also assigned.
546), ground (2343), ... "," labor (3112), unique (9821), ... "," mopile (2102), ... "," cherry blossom 1 (11111), orange 1 ( 9911) "," Sakura 2 (11112), Orange 2 (9912) ", and" Sakura 3 (11113) "are also assigned to each category. Further, the document addresses" SP1, SP2, SP3,... " “S01,
S02, S03,... "," CO1, CO2, CO
3, ... "," PL1, PL2, PL3, ... ",
"AR1, AR2, AR3, ..." and "EV1,
EV2, EV3,... "Are also assigned to each category.

Note that "Sakura 1", "Sakura 2" and "Sakura 3"
The first meaning (11111) and the second meaning (1111)
112) and the third meaning (11113). “Orange 1” and “Orange 2” indicate the first meaning (9911) and the second meaning (9912) of “orange”. For example, "orange 1" represents a plant orange, and "orange 2" represents an orange color. The reason that the meaning is used instead of the word itself in cases other than proper nouns is that the same word may have a plurality of meanings.

In step F15 in FIG. 4, for example, such a classification model is generated according to the manual classification operation by the user. Then, the classification model is registered as Step F16,
That is, it is recorded in the RAM 15 (or the HDD 34). The generation / registration of the classification model in this way means that the classification of the document has been performed.

Note that step F in FIG.
After the creation / registration of the classification model is performed as F15 and F16, the classification model is sequentially updated in accordance with automatic classification processing described later, editing of a classification item by a user, or manual classification operation. Become. When the classification model is updated, the update date and time are recorded in the classification model. FIG. 11 shows that the update date and time “19:05 on December 10, 1998
6 minutes and 10 seconds "are recorded.

4. 4. Automatic Classification Processing for Document Data 4-1 Processing Procedure In the document processing apparatus 1 of the present example, after the classification model is once created as described above, for example, the document data taken in from the outside by the communication unit 21 is automatically converted. The automatic classification process of classifying into is possible. That is, the automatic classification process described below is a process of, when the document processing apparatus 1 receives document data sent from the outside, classifying the document data with respect to a classification model. In this example, the automatic classification process described below may be performed each time one document is received, or may be performed each time a plurality of predetermined documents are received. The automatic classification process may be performed on all the documents received so far when the operation of opening is performed.

FIG. 12 shows the entire processing procedure as the automatic classification processing. Step F21 in FIG. 12 illustrates a document receiving process by the receiving unit 21 of the document processing apparatus 1. In this step F21, the receiving section 21 receives one or a plurality of documents transmitted via a communication line, for example. The receiving unit 21 sends the received document to the main body 10 of the document processing device. The control unit 11 stores the supplied document data or document data in the RAM 14 or the HDD 34.

Subsequently, the process proceeds to step F22, where the control unit 11
Creates an index for the document taken in step F21.

In step F23, the control unit 11 converts each indexed document based on the classification model into
Automatic classification into one of the categories of the classification model. Then, the control unit 11 causes the RAM 14 to store the result of the classification, for example. Details of the automatic classification will be described later.

In step F24, the control unit 11 updates the classification model based on the result of the automatic classification of the new document in step F23. Then, in step F25, the control unit 11 registers the classification model updated in step F24. For example, the classification model is stored in the RAM 14.

By the above-described processing of FIG. 12, automatic classification processing is performed so that the document data input to the document processing state 1 is classified on the classification model. That is, in this automatic classification process, an index is created for the received document, and after the automatic classification is performed, proper nouns, meanings,
The document address and the like are associated with a certain category on the classification model as shown in FIG. 11 (the classification model is updated).

The processes in steps F21 and F22 are the same as steps F11 and F12 in the above-described manual classification process. That is, as the index creation processing in step F22, the processing described with reference to FIGS. 5 to 8 is performed, and a repeated description thereof will be omitted. The update of the classification model in step F24 depends on the classification result of the automatic classification in step F23. Hereinafter, automatic classification in step F23 will be described in detail as processing different from the above-described manual classification processing.

4-2 Automatic Classification FIG. 13 shows the detailed processing of the automatic classification in step F23 in FIG. In step F61 of FIG.
Defines the number of intersections of the set of proper nouns included in the category Ci of the classification model and the set of proper nouns among the words extracted from the document received in step F21 and put into the index by P ( Ci). Then, the control unit 11 calculates the number P (Ci) thus calculated.
Is stored in the RAM 14.

In step F62, the control unit 11
Refers to the degree of association between all meanings included in the index of the document and all the meanings included in each category Ci by referring to the table of association between meanings shown in FIG. The total sum R (Ci) of the degrees is calculated. That is, the control unit 11 calculates the total sum R (Ci) of the degrees of association between all the meanings of words other than the proper noun in the classification model. The control unit 11 then calculates the total sum R (C
i) is stored in the RAM 14.

Here, the degree of association between meanings will be described. The degree of association between meanings may be calculated in advance for the meanings included in the electronic dictionary included in the document processing apparatus 1 by the processing in FIG. 14 and the result may be stored as shown in FIG. That is, if the control unit 11 executes the process of FIG. 14 once in advance, it can be used in the automatic classification process of FIG.

The processing of FIG. 15 which is executed in advance by the control unit 11 is as follows. First, in step F71, the control unit 11 uses the description of the meaning of a word in the electronic dictionary and creates a meaning network using this dictionary. That is, a meaning network is created from the reference relation between the explanation of each meaning in the dictionary and the meaning appearing in the explanation.
The internal structure of the network is described by tagging as described above. For example, the control unit 11 of the document processing apparatus sequentially reads meanings and descriptions of the electronic dictionary stored in the RAM 14 and creates a network. The control unit 14 causes the RAM 14 to store the meaning network created in this manner.

The above network is not only created by the control unit 11 of the document processing apparatus using a dictionary, but also by the receiving unit 21.
The recording / reproducing section 31 reproduces the data received from the outside from the recording medium 32. Further, the electronic dictionary can be obtained by receiving from the outside by the receiving unit 21 or by reproducing from the recording medium 32 by the recording / reproducing unit 31.

In Step F72, Step F7
The central activity value corresponding to each semantic element is spread on the semantic network created in step 1. By this activity diffusion, the central activity value corresponding to each meaning is given according to the internal structure by tagging given by the dictionary. The diffusion processing of the central activity value is the processing described with reference to FIG.

In Step F73, Step F7
One semantic Si constituting the semantic network created in 1 is selected, and in the subsequent step F74,
The initial value of the central activity value ei of the vocabulary element Ei corresponding to this meaning Si is changed, and the difference △ ei of the central activity value at this time is calculated.

Further, in step F75, a difference Δej of the central activation value ej of the element Ej corresponding to another meaning Sj corresponding to the difference Δei of the central activation value ei of the element Ei in step F74 is determined. In step F76, the difference Δe obtained in step F75
quotient ｊej obtained by dividing j by △ ei obtained in step F74
/ △ ei is the degree of association between meanings of the meaning Si with respect to the meaning sj.

In step F77, one meaning Si
It is determined whether or not the calculation of the degree of association between meanings has been completed for all pairs of the word and another meaning Sj. When the calculation of the relevance between meanings is not completed for all pairs of meanings, the process returns to step F73, and the calculation of the relevance between meanings is continued for the pairs for which the calculation of the relevance is not completed. In the loop from step F73 to step F77, the control unit 11 sets a necessary value to, for example, RA
M14 is sequentially read out, and the degree of association between meanings is calculated as described above. The control unit 11 causes the RAM 14 to sequentially store the calculated meaning-to-sense associations, for example. Then, when the calculation of the degree of inter-word meaning is completed for all the word-sense pairs, this series of processing is ended from step F77.

The calculation of the degree of association between meanings is performed by a certain one.
It can be said that, when the central activity value of one meaning is changed, the meaning in which the central activity value changes along with the meaning has a high degree of relevance. That is, if the central activity value of a certain meaning is changed in step F74, the central activity value of the related (linked) meaning changes accordingly, and if the degree of the change is examined, a certain meaning is obtained. The degree of relevance of each other meaning with respect to is understood. (The central activation value of an element Ei reflects the central activation value and the endpoint activation value of the linked element as described in the above description of the activity diffusion, and the element Ei endpoint activation value is updated. In addition, since the value is obtained from the sum of the end point activity value of the element Ei and the current center activity value, the larger the degree of association with the link destination is, the larger the amount of change of the center activity value is.) By performing the processing for all other meanings, it is possible to calculate the degree of association for all pairs (combinations) of meanings.

The calculated degree of association between meanings is defined between each meaning as shown in FIG. In the table of FIG. 15, the degree of association between meanings is normalized to take a value from 0 to 1. In this table, as an example, the degree of inter-semantic association between “computer”, “television”, and “VTR” is shown. The relevance between meanings of "computer" and "television" is 0.55, the relevance between meanings of "computer" and "VTR" is 0.25, and the relevance between meanings of "television" and "VTR" is 0.60. It is.

When the processing in step F62 in FIG. 13 is performed using the degree of association between the meanings calculated in advance as described above, the control unit 11 subsequently proceeds to step F63, where the inter-document classification for the category Ci is performed. Relevance Rel (C
i) is calculated as Rel (Ci) = m1P (Ci) + n1R (Ci). Here, the coefficients m1 and n1 are constants and represent the degree of contribution of the respective values to the inter-document-class relevance. The control unit 11 performs the calculation of the above expression using the number P (Ci) of common sets calculated in step F61 and the sum R (Ci) of the degree of association between meanings calculated in step F62, and calculates the degree of association between document classifications. Calculate Rel (Ci). The control unit 11 causes the RAM 14 to store the thus calculated inter-document-class relevance Rel (Ci).

The values of these coefficients m1 and n1 can be, for example, m1 = 10 and n1 = 1. The values of the coefficients m1 and n1 can also be estimated using a statistical method. That is, the control unit 11 determines the degree of association Rel between document classifications for a plurality of pairs of coefficients m and n.
Given (Ci), the coefficient can be obtained by optimization.

In Step F64, the control unit 11
Is the document class relevance Rel (C
When i) is the largest and the value of the degree of association Rel (Ci) between the document classifications exceeds a certain threshold, the documents are classified into the category Ci. That is, the control unit 11 creates the inter-document-class relevance for each of the plurality of categories, and when the maximum inter-document-class relevance exceeds the cabinet value, the control unit 11 assigns the document with the maximum inter-document-class relevance. Classify into category Ci. As a result, the document is automatically classified into a required category. If the maximum degree of association between document classifications does not exceed the threshold value, no document classification is performed.

The processing shown in FIG.
If the automatic classification in step F23 is performed,
The series of automatic classification is completed by updating and registering the classification model accordingly at 24 and F25. That is, the document data received by the document processing apparatus 1 is automatically classified, and the user can confirm the received document data in a required document classification area, for example, in the classification window 201 of FIG. Become.

5. Summary Creation Processing Next, processing for creating a summary sentence for document data will be described. As described above, the user can browse the text of the document by selecting the document and opening the browse window 301 as shown in FIG. For example, the viewing window 301 can be opened from the classification window 201 described with reference to FIG.

For example, when a browser button 202b is clicked while a certain document is selected in the classification window 201, a browsing window 301 in which the text of the selected document is displayed in the document display unit 303 as shown in FIG.
Is opened. When the entire text of the document cannot be displayed on the document display unit 303, a part of the document is displayed. At the time when the summary sentence has not been created, the summary display section 304 is blank as shown in FIG.

When the user clicks the summary creation button 306a in the browsing window 301, the document display unit 30
A summary sentence for the document displayed in 3 is created,
It is displayed on the summary display section 304 as shown in FIG. That is, the control unit 11 responds to the user's summary creation operation by
A summary sentence creation process as described below is performed, and after creation, control for displaying the summary sentence is performed. The process of creating an abstract from a document is performed based on the internal structure of the document by tagging. The summary sentence is generated according to the size of the summary display unit 304. The area of the text display unit 303 and the summary display unit 304 can be changed by moving the partition frame 312 by the user. In other words, the summary sentence is created in a size (document length) according to the size of the summary display unit 304 at the time when the summary creation is instructed.

FIG. 1 shows a summary creation process of the control unit 11, which is started when the summary creation button 306a is clicked.
FIG.

At step F81 of FIG.
Performs active diffusion. In this example, by adopting the central activity value obtained by the active diffusion as the importance,
Summarizes the document. That is, in a document given an internal structure by tagging, by performing active diffusion, a central activation value corresponding to the internal structure by tagging can be given to each element. The active diffusion process performed in step F81 is the same process as that described with reference to FIGS. 6 to 8, but the active diffusion is performed as described above.
This is a process of giving a high central activity value to an element related to an element having a high central activity value. That is, the active diffusion is an operation on the central activity value such that the central activity value is equal between the anaphor (co-reference) expression and its antecedent, and the central activity value attenuates otherwise. Since the central activity value is determined according to the internal structure by tagging, it can be used for analyzing a document in consideration of the internal structure by tagging.

Next, at step F82, the control unit 11
The size of the summary display section 304 of the browsing window 301 displayed on the display section 30, specifically, the summary display section 3
The maximum number of characters that can be displayed in 04 is set as ws. Further, the control unit 11 initializes a summary character string (an internal register holding a summary sentence) and sets an initial value s (0) = “”. The control unit 11 sets the maximum number of characters w
The s and the initial value s (0) of the character string s are recorded in the RAM 14.

In step F83, the control section 11 sets the count value i of a counter for counting the sequential creation of the skeleton of the digest to "1". Then, in step F84, the control unit 11 extracts a skeleton of a sentence having the ith highest average central activity value from the sentence based on the count value i of the counter. The average central activity value is the average of the central activity values of the elements constituting one sentence. Control unit 11
Is, for example, a character string s (i-1) recorded in the RAM 14.
Is read out, and the character string of the skeleton of the extracted sentence is added to this character string s (i-1) to obtain S (i). Then, the control unit 11 converts the character string s (i) thus obtained into R
Record in AM14. In the first time, since the character string s (i-1) has the initial value s (0), the skeleton of the sentence extracted this time is stored in the RAM 14 as the character string S (i).
When the process of step F84 is performed thereafter, the skeleton of the extracted sentence is changed to the character string S (i), and the character string S (i) up to that point (that is, the character string S (i-
1)). At the same time, the control unit 11 determines in this step F84 a list L (i) of the elements not included in the skeleton of the sentence in the order of the central activation value.
And the list L (i) is recorded in the RAM 14.

That is, in step F84, the summarization algorithm uses the result of activity diffusion to
The sentences are selected in descending order of the average central activity value, and the skeleton of the selected sentence is extracted. The skeleton of a sentence is composed of essential elements extracted from the sentence. Required elements can be the head of the element, the subject, the object (obj
ect), indirect object, owner (poss
An element having a relationship attribute of essor, cause, condition, or comparison, and an element directly included when the coordination structure is an essential element, constitute an essential element. Then, by connecting the essential elements of the sentence, a skeleton of the sentence is generated and added to the summary.

At step F85, the control section 11 determines whether or not the length of the character string s (i) is larger than the maximum number of characters ws of the summary display section 104 of the browsing window 301.
This step F85 is a determination process for creating a summary sentence according to the size of the summary display unit 304.

If the length of the character string s (i) has not reached the maximum number of characters ws, the control unit 11 executes the processing in step F
Proceed to 86. In step F86, the control unit 11 determines in the document the central activity value of the element of the sentence having the (i + 1) -th highest average central activity value and the highest central activity value of the list L (i) created in step F84. Compare the central activity values of the elements. That is, a sentence having the second highest average central activity value after the sentence adopted as the summary in step F84 (that is, a sentence which is a candidate to be added to the next summary sentence) and a sentence adopted as the summary in step F84 The central activity values of elements excluded from the summary as not scaffolds are compared.

In the process of step F86, the part to be added next to the character string as the summary sentence is set in the immediately preceding step F86.
In the sentence adopted in S84, the process is to determine whether to select from those not adopted as a skeleton or to select from other sentences.

The list L is higher than the central activity value of the element in the (i + 1) -th sentence having the highest average central activity value.
In the case where the highest central activity value in (i) is a higher central activity value, the portion to be added next to the character string as the summary sentence is set as the skeleton in the sentence adopted in the immediately preceding step F84. Try to choose from those that have not been adopted. For this reason, the process of the control unit 11 proceeds to step F88, in which the element having the highest central activity value in the list L (i) is stored in the character string S (i) stored at that time.
And a character string SS (i). At this time, the element added to the character string SS (i) is deleted from the list L (i). Then, in a step F89, the character string S
It is determined whether S (i) is greater than the maximum number of characters ws. If not, the process returns to step F86.

In step F86, if the (i + 1) -th element having the highest central activity value is higher than the highest central activity value in the list L (i), the summary sentence The part to be added next to the character string as is selected from a sentence different from the sentence adopted in step F84 immediately before that step F87.
, The count value i is incremented, and the process returns to step F84. That is, in step F86, (i + 1)
For the sentence having the second highest average central activity value, the skeleton is extracted in step F84 and added to the character string S (i).

As described above, in step F84 or step F88, while adding elements to the character string based on the element having the higher central activation value as the element or other element serving as the skeleton of the sentence in step F84 or step F88,
9, the character string S (i) or SS (i) is converted to the maximum number of characters ws
By comparing with, a character string close to the maximum number of characters ws but not exceeding the maximum number of characters ws is created.

For example, if the character string S (i) exceeds the maximum number of characters ws in step F85, the processing of the control unit 11 proceeds to step F90, and the character string S (i- 1) is a summary sentence. In other words, this means that the maximum number of characters ws has been exceeded by adding the skeleton of the sentence in step F84.
The character string S (i-1) before adding the skeleton is determined to be a character string that is close to the maximum number of characters ws but does not exceed the maximum number of characters ws, and is used as a summary sentence.

If the character string S (i) exceeds the maximum number of characters ws at step F85 when the character string S (i) is first generated at step F (i = 1) at step F85, Since the character string S (i-1) becomes the character string S (0) as the initial value set in step F82, a summary sentence could not be created substantially. This is because the size of the summary display unit 304 is too small.
The user may enlarge the area of the summary display unit 304 on the screen, and then click the summary creation button 306a again to start the processing in FIG.

If it is determined in step F85 that the character string S (i) does not exceed the maximum number of characters ws, as described above, the control unit 1
The process of step 1 proceeds to step F86, where a portion to be added to the next character string is determined. When the process proceeds to step F88 as described above, the character string SSi is
s is determined. Here, if the character string SSi exceeds the maximum number of characters ws, the process of the control unit 11 proceeds to step F91, and the character string S (i) before adding a certain element in the immediately preceding step F88 is used as a summary sentence. Will be. In other words, this means that the maximum number of characters ws has been exceeded by adding an element in step F88.
It is determined that (i) is a character string that is close to the maximum number of characters ws but does not exceed the maximum number of characters ws, and is used as a summary sentence.

By the above processing, a summary sentence suitable for the size of the summary display unit 304 at that time is created. The contents of the summary sentence include the skeleton of one or more sentences having a high average central activity value and elements having a high central activity value other than the skeleton. The summary sentence thus created is stored in the RAM 14 and displayed on the summary display unit 304 as shown in FIG.

When the user looks at the displayed summary and wants to see a more detailed summary, or wants to see a shorter summary, the viewing window 30 is displayed.
After increasing or decreasing the size (area) of the first summary display unit 304, the user may click the summary creation button 306a again. Then, a summary having a document length corresponding to the size of the summary display unit 304 at that time is created and displayed by the processing of FIG. 18 described above.

6. Read-aloud processing When the document processing apparatus 1 receives a document tagged as described above, the document processing apparatus 1 can display the body and the summary sentence to the user, and can also present the received document to the user by voice. . That is, the ROM 13
5 or an electronic document processing program recorded in the HDD 34 to activate a speech reading program,
The document can be read aloud through a series of steps as shown in FIG. In addition, at the time of reading out, an image of a character such as a person is displayed on the display unit 30 as the image of the speaker who is reading out, and depending on the document, reading out by a plurality of types of voices and character display corresponding thereto are performed. Have to be able to run. First, here, a simplified description of each step is given, and thereafter, each step is explained in detail using a specific document example.

As the processing of the control section 11, step F101 in FIG. 19 is the same document receiving / storing processing as step F11 in FIG. 4 (or step F21 in FIG. 12). As described above, the received document (tagged document) is subjected to manual or automatic classification processing. Similarly, in step S101, FIG. It is written. Here, the order of the classification process and the reading process and the like are not particularly defined as the processing procedure.

It should be noted that a document (reception document) to be subjected to the text-to-speech process needs to be provided with a tag necessary for performing speech synthesis, as described later. However, after receiving the tagged text, the document processing apparatus 1 can also create a document by newly adding a tag necessary for performing speech synthesis to the document. Hereinafter, the tagged document prepared by receiving or creating in this manner is referred to as a tag file.

In the document reading process of the document processing apparatus 1, a reading file is generated based on the tag file under the control of the CPU 13 in step F102. As will be described later, the reading file is generated by deriving attribute information for reading from a tag in the tag file and embedding the attribute information.

Subsequently, in step F103, under the control of the CPU 13, the document processing apparatus 1 performs a process suitable for the speech synthesis engine using the reading file.
The speech synthesis engine may be configured by hardware or may be realized by software. When the speech synthesis engine is realized by software, the application program is stored in the ROM 15 or the HDD 34 in advance.

Subsequently, the document processing apparatus 1 proceeds to step F104.
In, processing is performed according to an operation performed by a user using a user interface described later. Document processing device 1
Can read out a given document by performing such processing. Each of these steps will be described in detail below.

First, reception or creation of a tagged document in step F101 will be described. The document processing apparatus 1 receives, for example, a document (a document to which a tag required for performing speech synthesis has already been added) from the communication unit 21. Alternatively, the document processing apparatus 1 receives a tagged document, and newly adds a tag necessary for performing speech synthesis to the document to create the document.

As an illustrative example, it is assumed that the following tag file in which Japanese and English documents are tagged is received or created.

First, a Japanese document serving as a source of a tag file is as follows. "[Nice aging] / 8 cancer metastases can be suppressed! Cancer has been the leading cause of death in Japan for more than a decade.
Reed mortality is increasing with age. When thinking about the health of the elderly, you cannot avoid cancer.
Characterizing cancer is cell proliferation and metastasis. There are two types of human cells: "cancer genes", which act as accelerators in automobiles and rapidly grow cancer, and "cancer suppressing genes" that act as brakes. There is no problem if both are balanced. Loss of normal regulatory function and nonbreaking mutations in the cell initiate cancer growth. In the elderly, the mutation accumulates over the years, increasing the proportion of cells with cancerous conditions, leading to more cancers. By the way, without the other characteristic, metastasis, cancer doesn't have to be so afraid. This is because a complete cure is possible just by resection. Here is the importance of suppressing metastasis. This metastasis does not occur simply by increasing the number of cancer cells. Cancer cells dissolve proteins (proteins) between cells, create their own way, and enter blood vessels and lymph vessels. In recent years, it has been elucidated that complicated movements such as searching for a new “dwelling house” while circulating are performed. "

The following is an example of an English document serving as a tag file: "During its centennial year, The Wall Street Jour"
nal will report events of the past century that st
and as milestones of American business history.TH
REE COMPUTERS THAT CHANGED the face of personal co
mputing were Iaunched in 1977. That year the Apple
II, Commodore Pet and Tendy TRS came to market.T
he computers were crude by today's standerds.Appl
e ll owners, for example, had to use their televisi
on sets as screens and storeddata on audiocassette
s. "

When the document processing apparatus 1 receives a document with a tag attached to such a Japanese or English document,
As described with reference to FIG. 16 and FIG. 17 and the like, it is possible to display the main text or create and display a summary sentence.

Here, the above-mentioned Japanese or English documents are each constituted as a tag file as shown in FIG. 21 or FIG. As a tag file of a Japanese document, FIG. 21 (a) shows an excerpt of the heading “[nicely aging] / 8 cancer metastasis, suppressed !?”, and FIG. 21 (b) It shows an excerpt from the last paragraph in the document, "This metastasis, cancer cells are being elucidated in recent years." The remaining paragraphs have been omitted. It should be noted that the actual tag file is composed of one file from the heading to the last paragraph.

In the heading shown in FIG.
The tag <heading> indicates that this part is a heading. In this example, actually <header speaker = “sp1”>, which will be described later. The tag files shown in FIGS. 21A and 21B are basically attached with tags in the same manner as the tag file example used when the document data structure was described with reference to FIG. Although a detailed description of each of the above-described tags is omitted, tags required for performing speech synthesis are added to required locations.

For example, as a tag necessary for performing speech synthesis, first, as shown in the portion of “Example 1” in the figure,
Some are given when information indicating a reading pseudonym is given to the original document, such as “protein”.
In other words, in this case, the pronunciation = “nul
Reading attribute information “l” is described, and a tag for prohibiting partial reading of “(tanpaku)” is added. Also, the tags required to perform speech synthesis include:
As shown in the "Example 2" and "Example 3" in the figure, technical terms such as "lymphatic vessels" and difficult-to-read parts such as "dwelling house" that may be read aloud incorrectly There is. In other words, in this case, "Rinpakuda" or "Sumie"
In order to prevent the pronunciation from being read out, pronunciation attribute information indicating the pronunciation pseudonym of pronunciation = “Rinkan” and pronunciation = “Sumika” is described.

On the other hand, as a tag necessary for performing speech synthesis in the tag file shown in FIG. 22, as shown in a portion shown as “Example 4” in the figure, for the Roman numeral “II”, the pronunciation = “ Two ”reading attribute information is described. This is described in order to prevent "II" from being read out as "second" when it is desired to be read out as "two".

Although not shown in the tag files of FIGS. 21 and 22, for example, when a document includes a quote, a tag indicating that the sentence is a quote is added. . Further, when there is a question sentence in the document, for example, a tag indicating that the sentence is a question sentence is added to the tag file, though not shown.

Further, as a tag necessary for performing voice synthesis, a tag (voice type information) for specifying a speaker at the time of voice synthesis, that is, a type of a read-out voice to be output is added. The voice type here refers to, for example, a sound quality setting according to various person settings. For example, sound quality settings according to various person settings, such as a young woman, a young man, a boy, a girl, a middle-aged man, and a middle-aged woman, as speakers who read aloud. The speech synthesis engine is capable of outputting a read-aloud voice with a sound quality corresponding to each of such a plurality of person settings, and a tag for specifying a type of voice (speaker) is included in each part of the document, This specifies the person setting (speaker) at which the voice output is performed. For the sake of explanation, for example, codes “sp1”, “sp2”,.
In the tag, this code specifies the type of the voice to be read.

That is, in the case of FIG. 21, as shown in “Example 11”, the tag of <heading> described above is set to <heading speaker = “sp1”>. = “Sp1”> to </ headline>, it is instructed that the speaker at the time of reading out is performed by the character “sp1”. As shown as “Example 12,” FIG. 21B shows that the tag <paragraph> indicating the paragraph is <paragraph speaker = “s” here.
p2 ">, whereby <paragraph speaker =“ sp ”
In the range of 2 ″> to </ paragraph> (that is, the whole sentence of FIG. 21B), it is instructed that the speaker at the time of reading out is performed by the character “sp2”.

In FIG. 22, the tag (voice type information) for specifying the type (speaker) of the read-out voice includes:
As shown in “Example 13” to “Example 17”, the tag added to the head of the sentence is <speaker = “sp1”> or <speaker = “sp2”>. That is, in this example, the speaker is alternately designated by sp1 and sp2 for each sentence (<sentence> to </ sentence>). That is, it is specified that the sound quality of the read-out voice is switched alternately for each sentence.

In the example of FIG. 21, the speaker is specified in paragraph units, and in the example of FIG. 22, the speaker is specified in sentence units. However, this is merely an example for explanation, and the speaker is specified. The tag to be added can be arbitrarily added in units of sentences, units of paragraphs, units of sentences, and in some cases, units of phrases. In practice, it is preferable that a speaker is set for each appropriate portion according to the contents of the document.

In step F101 shown in FIG. 19, the document processing apparatus 1 receives or creates a document to which a tag necessary for performing speech synthesis has been added, for example, as in the above example. .

Next, the generation of a reading file in step F102 in FIG. 19 will be described. The document processing device 1 derives attribute information for reading out from a tag in the tag file, and generates a reading-out file by embedding the attribute information. Specifically, the document processing apparatus 1 finds tags indicating the beginning of a paragraph, a sentence, or a phrase of a document, and embeds attribute information for reading aloud in correspondence with these tags. When the document processing apparatus prepares the summary of the document, the head of the part included in the summary is found from the document, and the attribute information for increasing the volume at the time of reading out is embedded. Can be emphasized.

The document processing apparatus 1 generates a reading file as shown in FIG. 23 or 24 from the tag file shown in FIG. 21 or FIG. FIG. 23 (a)
(B) corresponds to the portion shown in FIGS. 21 (a) and (b). It goes without saying that the actual reading file is composed of one file from the above-mentioned heading portion to the last paragraph.

In the reading file shown in FIG. 23, attribute information of Com = Lang *** is embedded corresponding to the head of the document. This attribute information indicates the language describing the document. Here, the attribute information is Com = Lang = JPN, indicating that the language describing the document is Japanese. In the document processing apparatus, by referring to the attribute information, it is possible to select an appropriate speech synthesis engine corresponding to the language for each document.

In addition, attribute information such as Com = begin_p, Com = begin_s, and Com = begin_ph is embedded in this reading file at various places. These pieces of attribute information indicate the beginnings of paragraphs, sentences, and phrases of the document, respectively. The document processing device 1 identifies the beginning of these paragraphs, sentences, and phrases based on the tags in the tag file described above. In the reading file, for example, for portions where tags representing syntactic structures of the same level appear continuously, such as <adverb phrase><nounphrase> in the above-described tag file, the corresponding numbers are used. Are not combined, and one Com = begin_ph is embedded.

Further, the read-out file includes Com = be
Attribute information of Pau = 500, Pau = 100, and Pau = 50 is embedded corresponding to gin_p, Com = begin_s, and Com = begin_ph, respectively. These pieces of attribute information indicate that pause periods of 500 ms, 100 ms, and 50 ms are provided for reading out. That is, the document processing apparatus 1 provides pauses of 500 ms, 100 ms, and 50 ms at the beginning of a paragraph, a sentence, and a phrase of a sentence, respectively, so that the document is read out by the speech synthesis engine. Information. Note that these pieces of attribute information are embedded corresponding to Com = begin_p, Com = begin_s, and Com = begin_ph. Therefore, for example, a portion where tags representing the same level of syntactic structure appear continuously, such as <adverb phrase><nounphrase> in the tag file, is regarded as one phrase, and the corresponding number of Pau = One Pau = 50 is embedded together without being embedded. Also,
For example, Pau = *** is embedded in portions where tags representing different levels of syntactic structure appear continuously, such as <paragraph><sentence><nounphrase> in the tag file. Therefore, when reading out such a portion, the document processing apparatus 1 provides a pause period of 650 milliseconds obtained by adding the pause periods of the paragraphs, sentences, and phrases of the document, for example. in this way,
By providing a pause period corresponding to a paragraph, a sentence, and a phrase, the document processing apparatus 1 can perform reading without discomfort in consideration of a break between the paragraph, the sentence, and the phrase. This pause period does not need to be 600 ms, 100 ms, and 50 ms at the beginning of the paragraph, sentence, and phrase of the document, respectively, and can be changed as appropriate.

Further, in the reading-out file, “(protein)” is removed and the pronunciation = “rin” corresponding to the pronunciation attribute information of pronunciation = “null” described in the tag file. According to the reading attribute information of "kan" and pronunciation = "Sumika", "lymphatic vessels"
"Dwellers" are "Rinkan" and "Sumika", respectively.
Has been replaced by By embedding such reading attribute information, the document processing device 1 prevents reading errors due to inadequate dictionaries referenced by the speech synthesis engine.

In the reading file, "Chara = sp *" is used as the voice type information in the reading file in correspondence with the voice type information "speaker =" sp * "described in the tag file. Is embedded. That is, the head (Com = Lang = J) of the heading shown in FIG.
In the field next to PN, Chara = sp1 is embedded.
At the beginning of the paragraph shown in (b), Chara = sp2 is embedded.

In the reading file, attribute information for designating only this quotation to use another speech synthesis engine based on a tag indicating that the quotation is included in the document is included. It may be embedded. Further, attribute information for raising the intonation of the ending of the sentence may be embedded in the reading-out file based on a tag indicating that the sentence is a question sentence. Furthermore, in the reading-out file, if necessary, attribute information for converting a so-called "Dama-tona" style to a "Dama-tona" style can be embedded. In this case, instead of embedding such attribute information in the reading-out file, the document processing apparatus 1 converts the style of "Dana-tona" into the style of "Dan-Ma-tona" and converts the speech reading-out file. You may make it generate | occur | produce.

On the other hand, in the reading file shown in FIG. 24, attribute information of Com = Lang = ENG is embedded corresponding to the head of the document, indicating that the language describing the document is English. ing. The attribute information of Com = Vol = *** is embedded in the file for reading out.
This attribute information indicates the volume at the time of reading. For example,
Com = Vol = 0 indicates that the text is read out at the default volume of the document processing apparatus. Com = Vol = 80 indicates that the default volume is read out at a volume increased by 80%. Any Com = Vol = *** is valid until the next Com = Vol = ***. Further, in the reading file, "II" is replaced by "two" in accordance with the reading attribute information of pronunciation = "two" described in the tag file.

Further, also in this reading-out file, the speaker described in the tag file = “sp *”
Information that specifies the type of speech to be read, corresponding to the speech type information (speech type information in the file)
"Chara = sp *" is embedded. That is, Chara = sp1 or Chara = sp2 is embedded at the beginning of each sentence in FIG.

The document processing apparatus 1 generates such a reading file through a series of steps shown in FIG. First, the document processing apparatus 1 executes step F201.
The CPU 13 analyzes the received or created tag file. Here, the document processing apparatus 1 determines the language in which the document is described, and searches for the head of the paragraph, sentence, and phrase of the document, and the reading attribute information based on the tag. Subsequently, in step F202, the document processing apparatus 1 embeds attribute information Com = Lang = *** at the head of the document by the CPU 13 in accordance with the language in which the document is described.

Next, the document processing apparatus 1 executes step F203.
, The CPU 13 replaces the head of the paragraph, sentence and phrase of the document with the attribute information in the reading file. That is, the document processing apparatus 1 replaces <paragraph>, <sentence>, and <*** phrase> in the tag file with Com = begi
Replace with n_p, Com = begin_s and Com = begin_ph.

Further, the document processing apparatus 1 executes step F20.
In 4, the same level of syntactic structure appears by the CPU 13 and the same Com = begin _ ***
Combine into one Com = begin _ ***. Subsequently, in step F205, the document processing apparatus 1 embeds Pau = *** in correspondence with Com = begin _ *** by the CPU 13. That is, the document processing apparatus 1 embeds Pau = 500 before Com = begin_p, embeds Pau = 100 before Com = begin_s,
Embed Pau = 50 before gin_ph. And the document processing device 1
Is determined by the CPU 13 in step F206.
Replace with correct reading based on reading attribute information. That is, the document processing apparatus 1 removes “(protein)” based on the pronunciation attribute information “pronunciation =“ null ”, and based on the pronunciation attribute information“ pronunciation = “Rinkan” and pronunciation = “Sumika”. "Lymphatic vessels", "houses"
Substitute “Rinkan” and “Sumika”, respectively.

Further, the document processing apparatus 1 executes step F20.
7, the CPU 13 responds to the information “speaker = sp *” described in the tag file by using “Ch
ara = sp * ”.

In step F102 shown in FIG. 19, the document processing apparatus 1 automatically generates a reading file by performing the processing shown in FIG. The document processing device 1 causes the RAM 14 to store the generated reading file.

Next, the processing using the reading file in step F103 of FIG. 19 will be described.
The document processing device 1 uses the read-out file to
Under the control of the CPU 13, processing suitable for the speech synthesis engine stored in advance in the M15, the HDD 34, or the like is performed.
Specifically, the document processing apparatus 1 selects a speech synthesis engine to be used based on the attribute information of Com = Lang = *** embedded in the reading file. The speech synthesis engine is provided with an identifier according to a language or the like, and the information is recorded in the HDD 34 as, for example, an initialization file. The document processing device 1 refers to the initialization file and selects a speech synthesis engine having an identifier corresponding to the language. In addition, the speech synthesis engine is capable of outputting speech synthesis with sound quality as various speakers as described above,
Therefore, it is set so that each part (paragraph, sentence, or the like) is read aloud by voice corresponding to one or a plurality of speakers specified in the reading file.

The document processing device 1 converts Com = begin _ *** embedded in the file for reading out into a format suitable for the speech synthesis engine. For example, the document processing device 1
Mark Com = begin_ph with a number in the 10000s, such as Mark = 10000. Also mark Com = begin_s with a number in the 1000s, such as Mark = 1000, and Com = begin_p
Is marked with a number in the 100s, such as Mark = 100. This means that the heads of <phrase>, <sentence>, and <paragraph> are indicated by numbers in the 10000s, 1000s, and 100s, respectively.
The head of <sentence> and <paragraph> can be identified. Furthermore, in the reading file, the volume attribute information is
Since the volume is represented by the default volume percentage such as Voll = ***, the document processing apparatus 1 converts the percentage information into absolute value information based on the attribute information and obtains the absolute value information.

In step F103 shown in FIG. 19, the document processing apparatus 1 performs the processing using such a file for reading out, so that the speech synthesis engine can read out the document for reading out the file. Format.

Next, the operation using the user interface in step F104 of FIG. 19 will be described. The document processing device 1 activates the speech synthesis engine when the user operates the mouse or the like of the input unit 20 and clicks on the reading button 306c shown in FIG. 16 or FIG. Then, the document processing apparatus 1
A reading window 401 for a user interface as shown in FIG. 5 is displayed on the display unit 30.

[0176] As shown in the figure, the read-out window 401 has a play button 42 for reading out a document.
0 and a stop button 421 for stopping reading out
And pause button 4 to pause reading aloud
22. In addition, the reading window 401
Are a cue button 411, a fast rewind button 412, and a fast forward button 413 for cueing, fast rewinding, and fast forwarding in sentence units, a cueing button 414 for cueing, fast rewinding, and fast forwarding in paragraph units, and a fast rewinding button. A button 415 and a fast forward button 416, and a cue button 417 and a fast reverse button 4 for cueing, fast rewinding, and fast forwarding in phrase units
18 and a fast forward button 419.

Further, the reading window 401 is provided with a selection switch 4 for selecting whether to read the whole sentence or the summary sentence created as described above.
23,424.

Further, an image display area 403 for displaying an image as a speaker at the time of reading out is set, and an image of a person who is reading out is displayed. In addition, a telop display area 402 is provided so that characters can be displayed in telop corresponding to the reading voice.

Although not shown here, for example, a button for increasing / decreasing the voice, a button for increasing / decreasing the reading speed, and the like may be provided.

The document processing apparatus 1 performs a reading operation by the speech synthesis engine in response to the user clicking / selecting these various buttons / switches by operating the mouse or the like of the input unit 20, for example. For example, the document processing device 1
Starts reading the document by clicking the play button 420 by the user. Specifically, the control unit 1
1 supplies the audio signal generated by the audio synthesis processing to the audio output unit 33 and outputs it as audio. In addition, the document processing apparatus 1 stops or pauses the reading process when the stop button 421 and the pause button 422 are clicked.

Further, when the user presses the cue button 411 during the reading, the user jumps to the head of the currently read sentence and reads the sentence again. Cue button 4
In the same manner, for 14 and 417, the program jumps to the beginning of the currently read paragraph or phrase, and is read again. The control unit 11 controls the cue buttons 411, 414,
Regarding the operation 417, the jump destination is recognized by the above-mentioned marking. In other words, when the cue button 411 relating to a sentence is operated, the control unit 11 traces the sentence back from the middle position of the sentence currently being read out.
Search for 000's mark. When a mark in the 1000's range is detected, reading out is resumed from there. In the case of paragraphs and phrases,
Reading is resumed by looking for a mark in the 10000s range. By such processing, for example, it is possible to respond to a request that a user wants to repeatedly reproduce a desired portion in a document.

The document processing device 1 executes the processing in step F1 in FIG.
At 04, in response to the user performing an operation using the user interface in the reading window 401, the text-to-speech engine reads the document. In this manner, the document processing apparatus 1 can read out a desired document by the speech synthesis engine without any discomfort.

Further, at the time of reading, the image display unit 4
With the sound quality according to the speaker image displayed in 03,
Reading is performed. In other words, as described above, for each part (paragraph, sentence, etc.) in the document, a reading voice is output as the voice of the speaker specified by the tag, and the image display unit is performed according to the type of the speaker performing the reading. The display of 403 is switched. A description will be given of such a switching process between the reading voice and the display image.

It is assumed that the type of voice specified as Chara = sp1 is a young woman and the type of voice specified as Chara = sp2 is a young man. As an example of the operation when reading out the reading file in FIG. 23, in this case, the heading “Nicely Aging /
Since "8 cancer metastasis can be suppressed" specifies a young woman as Chara = sp1, the speech synthesis engine generates a reading voice signal in the sound quality (voice quality) set as the young woman. Therefore, the reading voice as the voice of the young woman is output to the user, and the image window of the young woman is displayed on the image display section 403 as the reading window 401 on the display section 30 as shown in FIG.

In the paragraph shown in FIG.
Since a young man is specified as Chara = sp2, the speech synthesis engine generates a reading voice signal in the sound quality (voice quality) set as this young man. Therefore, the read-aloud voice as the voice of the young man is output to the user, and the read-aloud window 401 on the display unit 30 displays the image image of the young man on the image display unit 403 as shown in FIG.

Such a display operation is performed by the control unit 11 shown in FIG.
7 is realized. FIG. 27 shows the process when the user instructs the start of the reading as step F104 in FIG. The control unit 11 causes the speech synthesis engine to start outputting the read-aloud voice as Step F301 in FIG. At this time, the speech synthesis engine reads out each paragraph, sentence, and phrase according to the pause time described in the reading-out file, and specifies the speaker (Ch
The output sound quality is switched according to ara = sp *).

As display control by the control unit 11, in step F302, corresponding character image data, for example, the above-mentioned image of a young woman, etc., is displayed in accordance with the speaker's designation (Chara = sp *) in the reading-out file. Data is read from the memory 16 and displayed on the image display unit 403 in the reading window 401.

Depending on the voice synthesis engine, Chara = sp *
Until the reading of the part (paragraph, sentence, etc.) specified by is completed, the voice output with the specified voice quality is performed, so that the voice output with a certain voice quality continues for a required period. At this time, as display control of the control unit 11, in step F303, control is performed to change the expression of the speaker image. This will be described later.

When the voice quality output by the speech synthesis engine is switched, that is, when the speech is read, the portion (paragraph,
When the voice quality to be output is switched after the sentence is reached, the display control of the control unit 11 is performed in steps F304 to F3.
Return to 02. That is, the control unit 11 reads out the character image data corresponding to the Chara = sp *, for example, the above-mentioned image image of a young man from the image memory 16 and displays it on the image display unit 403 in the reading-out window 401.

The above processing is performed until the completion of the reading is detected in step F305, that is, all the speech synthesis output based on the reading file by the speech synthesis engine is completed, or the user performs the operation of stopping the reading. Until this is detected, the image of the speaker image corresponding to the reading voice is displayed in the reading window 401 by being executed. That is, as described above, in the example of FIG.
Speaking is performed with the voice quality of the young woman with the display state of 5, and in the paragraph section of FIG. 23B, the reading is performed with the voice quality of the young man with the display state of FIG.

When the reading file of FIG. 24 is read out, the voices of the young woman and the young man are switched one sentence at a time, and the display state is correspondingly shown in FIGS. The display of the speaker image such as 26 is switched.

By the way, the expression control of the speaker in step F303 of FIG. 27 is executed, for example, as follows. For example, as an image of a young man corresponding to Chara = sp2, a plurality of image data are prepared in the image memory 16 as shown in FIGS. And the control unit 11
As Step F303, every predetermined time (for example, 0.1 to
Every one second), the image data of FIG. 28 is switched to generate an image as the reading-out window 401. Then, during a period in which the reading is performed by the voice of the young man, the image display unit 4 of the reading window 401 is displayed.
In 03, an image can be presented to the user as if the young man were actually talking while changing facial expressions.
That is, by switching a plurality of still images in a short time to realize a pseudo moving image, the display at the time of reading out can be made more realistic and interesting. In particular, the reading voice interval, that is, the period defined by Pau = ***,
An image in which the mouth is closed as shown in FIG. 28 (c) is displayed, and a period during which the reading voice is output is shown in FIGS.
By switching between (c) and (d), a display more suitable for the reading voice (that is, a display as if actually speaking) becomes possible.

FIG. 29 shows an example of processing for controlling such a displayed expression of a speaker. In the display control process of FIG. 27, the period during which the sound is output with the sound quality as a certain speaker (the period until the speaker is switched) is determined by step F
The expression control of the speaker image as 302 is performed, and the control unit 11 performs the processing of FIG. 29 as the expression control. First, in step F401, the control unit 11 determines whether or not audio is being output. Here, the audio output is in progress.
The period during which the synthesized speech is actually output, that is, <
This is the period during which the portions marked as "paragraph", "sentence", and "<**** phrase" are read out. And the period during which the voice is not being output here is Pau = 500, Pau = 100,
According to Pau = 50, this refers to a period during which the audio output is temporarily stopped, that is, a period in which the output audio is a break between paragraphs, sentences, and phrases.

During the audio output, the process of the control section 11 proceeds to step F402, and the display of the current image data is, for example, 0.
It is determined whether or not five seconds have elapsed (of course, 0.5 seconds).
Seconds). If the time has not elapsed, the process returns to step F401. If the time has elapsed, the process proceeds to step F403 to switch the image data to be displayed to another image data of the same speaker.

During the period in which it is determined that the voice is not being output in step F401, the control unit 11 determines in step F404 that the image data to be displayed is the image data of the same speaker in the non-speech state (ie, the mouth). Is closed image data).

For example, by the above processing,
For example, during the period in which the voice is read out by the voice of a young man, the image display unit 403 shown in FIGS.
By switching (c) and (d) every 0.5 seconds,
The image actually looks like a young man talking,
At the moment of a break between a paragraph, a sentence, and a phrase, an image with the mouth closed as shown in FIG. 28C is displayed. Therefore, a display as if actually speaking is realized according to the reading voice.

Although the above description is an example of displaying a pseudo moving image, it is not a pseudo moving image but an actual moving image video signal of about several seconds is stored in the image memory 16 and is reproduced repeatedly. The speaker can be expressed more realistically. It is also conceivable to display a young man or the like as a still image instead of a moving image or a pseudo moving image. In this case, the processing load on the control unit 11 can be reduced.

By the way, the image data (still images or a plurality of still images or moving images for the pseudo moving image) as the speaker images are stored in the document processing apparatus 1 in advance, and the communication unit 21 and the The data may be attached to the document data obtained from the recording medium 32 or the like so that it can be used. Of course, independently of the document data, it is also conceivable that the data is provided via a communication line or fetched from the recording medium 32. Alternatively, images of various speakers may be recorded in the HDD 34, and necessary image data may be fetched into the RAM 14 and used for display control when reading out. In addition, the user can enter any image,
You may enable it to be registered or changed as a speaker image. Further, the data setting as the speaker described above is not limited to a person, and may be any setting such as an animal, a machine, or a fantasy creature.

Here, display examples different from those in FIGS. 25 and 26 are shown in FIGS. In FIGS. 25 and 26, the speaker displayed on the image display unit 403 is switched according to the reading voice, but the speaker specified in the currently executing reading file is always displayed. You may do so. For example, in each of the above examples of the reading file, two characters, Chara = sp1 and Chara = sp2, are specified. Therefore, the period during which these reading files are being read is shown in FIGS.
, Each speaker is always displayed as the image display units 403a and 403b. The period in which the reading of the portion where Chara = sp1 is being executed is shown in FIG.
As shown in FIG. 27, the facial expression control is performed on the image of the young woman on the image display unit 403a as step F303 in FIG. 27, during which the image of the young man on the image display unit 403b is a still image, or the image is slightly darkened. And other processing. Conversely, during the period in which the reading of the portion where Char = sp2 is being executed, the image of the young man on the image display unit 403b as shown in FIG.
Expression control is performed as 303, during which the image of the young woman on the image display unit 403a is subjected to processing such as a still image or a slightly darker image.

[0200] In this way, the user can find out which two characters appear in the reading file (document data) currently being read, and which of those characters is currently talking. Can be accurately recognized.

In the above description, an example in which two characters (speakers) are specified for reading out one piece of document data and reading out is performed. Of course, three or more characters may be specified. , According to Chara = sp *,
The voice output based on the voice quality of a predetermined speaker and the display of an image of the speaker are performed.

As described above, the reading process of this example, that is, voice output and display as a character of each speaker is performed based on the tag specifying the speaker, thereby presenting a document which is easy for the user to understand and interesting. Can be realized. In the above example, explanation text such as news was explained. However, in actuality, in document data including a conversation text,
Realize presentations that make each character have a conversation, or change the speaker in the basic explanation part and the detailed explanation part with the document data as the explanation, and explain the document data for presentation by sharing each speaker The application range is wide, such as realizing the presentation that performs.

In the above-described reading operation of the present example, the document to be read may be the text of the document data or a summary sentence. Whether the text or the summary is selected by clicking the selection switches 423 and 424, in any case, the tag file as the text or the summary is read in steps F102 and F102 in FIG.
By performing the process of 103, the text-to-speech process by the above-described speech synthesis engine becomes possible.

In this example, the reading file is generated from the received or created tag file.
Instead of generating such a reading file, reading may be directly performed based on the tag file.
In this case, after receiving or creating the tag file, the document processing device 1 uses a speech synthesis engine and, based on tags indicating paragraphs, sentences, and phrases attached to the tag file,
A certain pause is provided at the beginning of a paragraph, sentence or phrase to read out. Further, based on the designation of the speaker in the tag file, the speaker for voice and display is selected.
By doing so, the document processing apparatus 1 can directly read out aloud based on the tag file without generating a voice reading out file.

Although the embodiment of the present invention has been described above, this is merely an example, and various configurations and processing examples of the document processing apparatus 1 can be considered. For example, the document processing apparatus 1 can be realized by a stationary or portable general-purpose personal computer, or may be a document processing apparatus as a dedicated device. There are various specific examples of devices that constitute the document processing apparatus 1. For example, the document processing apparatus 1
As an example of the input unit 20 in, not only a keyboard and a mouse but also other devices such as a tablet, a light pen, a wireless commander device using infrared rays and the like can be considered.

Further, in the embodiment, the description has been made assuming that the document is transmitted from the outside to the communication section 22 via a telephone line or the like, but the present invention is not limited to this. For example, the present invention can be applied to a case where a document is transmitted via a satellite or the like.
Or a document may be written in the ROM 15 in advance.

In the embodiment, an example of a method of tagging a document has been described. However, it is needless to say that the present invention is not limited to this tagging method. Furthermore, in the above-described embodiment, Japanese and English sentences have been exemplified, but it goes without saying that the present invention is not limited to these languages. As described above, it goes without saying that the present invention can be appropriately changed without departing from the spirit of the present invention.

Further, in the present invention, a disk-shaped recording medium or a tape-shaped recording medium in which the electronic document processing program including the above-described reading process and the display process at the time of reading is written is provided as the recording medium 32. Can be easily realized. That is, the recording medium of the present invention can be realized. The recording medium of the present invention can be similarly used as the HDD 34 shown in FIG. According to such a recording medium 32, a program for realizing the above-described document processing method can be provided, and the document processing apparatus of the present invention can be easily realized using, for example, a general-purpose personal computer. Of course, the recording medium 32 may be an optical disk, a magneto-optical disk, a magnetic tape, a memory card such as a flash memory, a memory chip, or the like, in addition to the floppy disk. Furthermore, the program for realizing the document processing method of the present invention can be provided also via a network communication such as the Internet. Therefore, the present invention can be applied to a program server or a recording medium in a communication process. Things.

[0209]

As will be understood from the above description, the present invention has the following effects. That is, according to the document processing apparatus and the document processing method of the present invention, for example, a plurality of types of voices respectively corresponding to a plurality of persons or the like are selectively used to generate a voice signal serving as a reading voice for electronic document data. The speaker image signal is output according to the type of the generated audio signal. As a result, the electronic document data can be read aloud, and a character image or the like as the speaker is displayed, so that a user-friendly and interesting document presentation mode can be realized. Further, by sequentially outputting a plurality of types of voices and speaker images corresponding thereto, a reading operation more suitable for the contents of the electronic document can be realized. For example, the speaker can be switched in the middle of the document data of the explanatory sentence, or the presentation in the conversation state can be actually realized in the document data as the conversation sentence. As a result, the user can easily understand the contents of the document data and the presentation mode is significantly improved.

The type of the generated audio signal is determined based on the audio type information (tag) added to the electronic document data, so that reading and displaying as intended by the creator of the electronic document are performed. (Such as setting of persons in these cases) can be made, and more appropriate presentation to the user can be made. Furthermore, as the speaker image, a still image signal, a moving image signal, or a pseudo moving image signal corresponding to the type of the output audio signal can be displayed accurately in accordance with the reading voice.

Further, according to the recording medium of the present invention, a program for realizing the document processing method of the present invention can be provided, and the document processing apparatus of the present invention can be easily realized using, for example, a general-purpose personal computer. Become like Thus, the general user can easily enjoy the above effect.

[Brief description of the drawings]

FIG. 1 is a block diagram of a document processing apparatus according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a document structure used in the embodiment.

FIG. 3 is an explanatory diagram of a window for displaying a sentence structure according to the embodiment;

FIG. 4 is a flowchart of a manual classification process according to the embodiment.

FIG. 5 is a flowchart of an index creation process according to the embodiment.

FIG. 6 is an explanatory diagram of an activation value of an element according to the embodiment.

FIG. 7 is a flowchart of an active diffusion process according to the embodiment.

FIG. 8 is a flowchart of a central activation value update process according to the embodiment.

FIG. 9 is an explanatory diagram of a classification window according to the embodiment.

FIG. 10 is an explanatory diagram of a browsing window according to the embodiment.

FIG. 11 is an explanatory diagram of a classification model according to the embodiment.

FIG. 12 is a flowchart of an automatic classification process according to the embodiment.

FIG. 13 is a flowchart of automatic classification according to the embodiment.

FIG. 14 is a flowchart of a word meaning association degree calculation process according to the embodiment;

FIG. 15 is an explanatory diagram of a degree of association between meanings according to the embodiment.

FIG. 16 is an explanatory diagram of a display example of a browsing window according to the embodiment.

FIG. 17 is an explanatory diagram of a display example including a summary of a browsing window according to the embodiment;

FIG. 18 is a flowchart of a summary creation process according to the embodiment.

FIG. 19 is a flowchart of a document reading process according to the embodiment.

FIG. 20 is a flowchart of a reading file generation process according to the embodiment;

FIG. 21 is an explanatory diagram of an example of a tag file according to the embodiment;

FIG. 22 is an explanatory diagram of an example of a tag file according to the embodiment;

FIG. 23 is an explanatory diagram of an example of a reading file according to the embodiment;

FIG. 24 is an explanatory diagram of an example of a reading file according to the embodiment;

FIG. 25 is an explanatory diagram of an example of a reading window according to the embodiment;

FIG. 26 is an explanatory diagram of an example of a reading window according to the embodiment;

FIG. 27 is a flowchart of a reading process according to the embodiment.

FIG. 28 is an explanatory diagram of an example of a pseudo moving image according to the embodiment;

FIG. 29 is a flowchart of expression control according to the embodiment.

FIG. 30 is an explanatory diagram of an example of a reading window according to the embodiment;

FIG. 31 is an explanatory diagram of an example of a reading window according to the embodiment;

[Explanation of symbols]

1 Document processing apparatus, 10 main body, 11 control unit, 12
Interface, 13 CPU, 14 RAM, 15
ROM, 16 image memory, 20 input unit, 21 communication unit, 22 portable terminal reader, 30 display unit, 31 recording / reproducing unit, 32 recording medium, 33 audio output unit, 34 H
DD, 50 mobile terminal, 51 display unit, 52 operation unit,
53 reader terminal, 54 control unit, 55 document memory,
56 receiving unit, 57 display drive, 59 audio output unit,
60 headphone terminal, 101 window, 201
Classification window, 301 reading window, 303 document display unit, 304 summary display unit, 401 reading window

────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] July 7, 1999 (July 7, 1999)

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name]

[Correction method] Change

[Correction contents]

When the character string S (i) exceeds the maximum number of characters ws in step F85 when the character string S (i) is first generated in step F84 (i = 1) at step F85, Since the character string S (i-1) becomes the character string S (0) as the initial value set in step F82, a summary sentence could not be created substantially. This is because the size of the summary display unit 304 is too small. Therefore, the user enlarges the area of the summary display unit 304 on the screen, and then clicks the summary creation button 306a again to return to FIG. What is necessary is just to make a process start.

[Procedure amendment 2]

[Document name to be amended] Statement

[Correction target item name]

[Correction method] Change

[Correction contents]

If it is determined in step F85 that the character string S (i) does not exceed the maximum number of characters ws, as described above, the control unit 1
The process of step 1 proceeds to step F86, where a portion to be added to the next character string is determined. When the process proceeds to step F89 as described above, it is determined whether or not the character string SS (i) has exceeded the maximum number of characters ws. Where the string SS
If (i) exceeds the maximum number of characters ws, the process of the control unit 11 proceeds to step F91, and the character string S (i) before adding a certain element in step F88 immediately before is used as a summary sentence. Become. In other words, this means that the maximum number of characters ws has been exceeded by adding the element in step F88. Therefore, the character string S (i) before the addition of the element is close to the maximum number of characters ws but the maximum number of characters ws Is determined to be a character string that does not exceed, and is used as a summary sentence.

Claims (13)

[Claims] 1. A read-out voice generating unit capable of generating a voice signal serving as a read-out voice for electronic document data by selectively using a plurality of types of voices, and a voice generated by the read-out voice generating unit. A document processing apparatus, comprising: image generating means for outputting a specific type of speaker image signal according to a type of a signal. 2. A display unit, a voice output unit, and a type of a voice signal generated by the voice output unit, wherein the voice signal generated by the voice output unit is output from the voice output unit. And control means for controlling a display operation on the display means using a speaker image signal output from the image generation means in response to the control signal. Document processing device as described. 3. The document processing apparatus according to claim 1, wherein said reading voice generating means selects a type of a generated voice signal based on voice type information added to the electronic document data. . 4. The method according to claim 1, wherein the image generating unit outputs a still image signal or a moving image as a speaker image signal corresponding to the type during a period in which the reading voice generating unit outputs a certain type of audio signal. The document processing apparatus according to claim 1, wherein the document processing apparatus outputs a signal or a pseudo moving image signal. 5. An audio signal which is used to select one of a plurality of types of voices based on voice type information added to the electronic document data, and to use the selected voice as a reading voice for the electronic document data. A voice generating means capable of generating a voice signal, and selecting one of the plurality of types of speaker images based on the voice type information, and outputting an image signal based on the selected speaker image. Means, comprising: a document processing apparatus. 6. A read-out voice generating step of generating a read-out voice signal which is a read-out voice in which a plurality of types of voices are selectively used for electronic document data and outputting the read-out voice signal from a voice output unit; A specific type of speaker image signal is output according to the type of audio being performed,
A document processing method, comprising the steps of: generating an image to be displayed on a display unit; 7. The document processing method according to claim 6, wherein in the reading voice generating procedure, a type of the generated voice signal is selected based on voice type information added to the electronic document data. . 8. In the image generation procedure, during a period in which a certain type of audio signal is output by the reading voice generation procedure, a speaker image signal corresponding to the type is output.
7. The document processing method according to claim 6, wherein a still image signal, a moving image signal, or a pseudo moving image signal is output. 9. An audio signal that selects one audio from a plurality of types of audio based on audio type information added to the electronic document data, and uses the selected audio as a read-out audio for the electronic document data. And a read-out voice generation procedure to be output from a voice output unit, based on the voice type information, select one speaker image from a plurality of types of speaker images, and generate an image signal based on the selected speaker image. An image generation procedure for outputting and displaying the image on a display unit. 10. A read-out voice generating procedure for generating a voice signal as read-out voice by selectively using a plurality of types of voices for electronic document data and outputting the voice signal from a voice output unit; A specific type of speaker image signal is output according to the type of audio being performed,
An image generation procedure to be displayed on a display unit, and an operation control program having the following. 11. The read-out voice generating step includes the steps of: generating a read-out voice based on voice type information added to the electronic document data;
2. An operation control program for selecting a type of an audio signal to be generated is recorded.
0. The recording medium according to 0. 12. In the image generation procedure, during a period in which a certain type of audio signal is output by the reading voice generation procedure, a still image signal or a moving image as a speaker image signal corresponding to the type. The recording medium according to claim 10, wherein an operation control program for outputting a signal or a pseudo moving image signal is recorded. 13. An audio signal which selects one audio from a plurality of types of audio based on audio type information added to the electronic document data, and uses the selected audio as a read-out audio for the electronic document data. And a read-out voice generation procedure to be output from a voice output unit, based on the voice type information, select one speaker image from a plurality of types of speaker images, and generate an image signal based on the selected speaker image. An image generation procedure for outputting and displaying the image on a display unit.