JPH0619965A - Natural language processor - Google Patents

Abstract

PURPOSE:To easily check the original intention of an input sentence by correspondingly storing an abbreviated name and a regular name, interpolating the character string of the abbreviated name in the input sentence by a regular character string and displaying the interpolated character string. CONSTITUTION:When abbreviated information exists in an input sentence in a process for analyzing the input sentence by syntax analyzing processing S30 and meaning analyzing processing S40, an abbreviatable information storing part S120 is referred to. Abbreviated character strings and their corresponding regular character strings are stored in the storing part S120 by a table format, and when an abbreviated character string is searched, a regular character string corresponding to the searched character string can be extracted. Information interpolated from the analytical result obtained by the processing S40 by using information obtained from an abbreviated expression interpolating part S50 and the contents of an ambiguous part temporarily determined by the device are added to the input sentece and displayed. Thereby when a character string with an abbreviated name is included in a sentence inputted for retrieval, the original purpose of the input sentence can easily be confirmed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a natural language processing device, and more particularly to a natural language processing device for retrieving information by a sentence input in natural language.

[0002]

2. Description of the Related Art In recent years, research and development of natural language processing devices for searching information by sentences in natural language have been conducted.
Here, the flow of processing in the conventional natural language processing apparatus will be described with reference to FIG.

In FIG. 9, in step S401, a sentence in natural language is received as an input. Step S
In 402, the input natural language sentence is decomposed into morphemes, which are basic units, and information such as a part of speech and an inflection is added. In the next step S403, the syntactic structure of the input sentence is analyzed from the result of the morphological analysis. Step S4
In 04, the semantic relation of the input sentence is analyzed based on the information obtained by the analysis.

In step S405, in the process of analyzing the input sentence by the syntax analysis process of step S403 and the semantic analysis process of step S404, if there is information omitted in the input sentence, the information is complemented. (Abbreviated complement processing). In step S406,
By using the information from this abbreviation complementation process, from the semantic analysis result obtained from the semantic analysis process, the information that is insufficient for searching the database or the ambiguous part in the input sentence is determined, and if necessary, The user is asked a question again, and the process returns to step S401.

When it is determined that the information is sufficient for searching the database, the search condition is set to the next step S407.
Hand over to. In step S407, the database 411 is searched according to the search condition determined by the question content determination process 406. Then, the search result is displayed in the next step S408. Note that this step S40
In 8, the search result is displayed to the user, and the information displayed in the dialogue history holding unit 409 is delivered. Then, in order to end the information retrieval or continuously repeat the question response, the process returns to the input sentence reception process of step S401.

FIG. 10 shows an example of a dialogue of this type of natural language processing apparatus. It is an example of a dialogue of information guidance of the Shinkansen timetable. In the figure, U1 to U13 are user input sentences. S1 to S13 are message outputs from the device.

In this example, the user inquires about the Shinkansen timetable for going to Osaka at U1. From S2 to U4, supplement the labor saving of the question sentence of U1 (tomorrow is September 26th) and resolve ambiguous parts (Osaka is Shin-Osaka. 10 o'clock and 10 am That is). S5 is the display of the first search result.

U5 is a further question for that search result. In subsequent S6 to U7, the ambiguity of the input sentence is resolved again. S8 is the display of the second search result.

U9 is a further question for the search results of S8. This question is about the company that runs the buffet on the train. From S9 to U11, the labor saving information is confirmed, and at S12, the trains are narrowed down.

U13 and S13 are dialogues for ending.

[0011]

[Problems to be solved by the invention] However,
In the above conventional example, an information search is performed without showing the information complemented by the device to the user, or if there is ambiguity in the information to be complemented, in order to narrow it down, further There were some drawbacks, such as repeating question and answer.

(1) The user cannot easily confirm the supplemented information.

(2) When some further question-and-answers for resolving the ambiguity of the complemented information overlap, which part of the complemented information is forgotten.

(3) When there are several types of information that can be complemented by the device and there is a difference in the confirmation operation of those information, it is difficult to distinguish them.

The present invention has been made in view of the above prior art, and when there is a character string of abbreviated name in a sentence input for retrieval, it is possible to easily confirm the original intention of the input sentence. The present invention aims to provide a natural language processor.

[0016]

In order to solve this problem, the natural language processing apparatus of the present invention has the following configuration. That is, in a natural language processing device that searches information by a sentence input in natural language, a storage unit that stores a simplified name and a canonical name in association with each other, an input unit that inputs a sentence to be searched, and an input unit Determining means for determining whether or not there is a character string of the abbreviated name stored in the storage means in the character string forming the sentence, and the character string when it is determined that the character string of the abbreviated name is present On the other hand, it comprises a complementing means for complementing the regular character string stored in the storage means, and a display means for displaying the sentence complemented by the complementing means.

[0017]

In the structure of the present invention, a character string used as an abbreviated name and a regular character string corresponding to this character string are stored in the storage means. Then, when the user's input for the search includes a character string of abbreviated name, it is complemented with the corresponding regular character string, and the completed sentence is displayed for confirmation.

[0018]

Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

FIG. 8 is a block diagram showing the structure of an apparatus according to an embodiment of the present invention.

In the figure, 1 is a CPU that controls the entire apparatus. Reference numeral 2 is a ROM that stores the operation processing procedure (program) of the CPU 1 or character fonts.
Reference numeral 3 is a RAM used as a work area of the CPU 1. Reference numeral 4 is a keyboard for inputting various information. Reference numeral 5 is a CRT for displaying various messages, and 6 is a video RAM (VRAM) for developing an image to be displayed on the CRT 5. Reference numeral 7 is a dictionary / knowledge base used when analyzing morphemes of natural language input from the keyboard 4 and interpreting meanings. Reference numeral 8 is a database in which various information is structurally stored and managed.

The operation processing procedure of the apparatus of the embodiment having the above configuration will be described with reference to the flowchart of FIG. It should be noted that all the processes in the following description are performed by the CPU 1.

First, in step S10, a sentence input from the keyboard 4 in a natural language format is received. In step S20, the received input sentence is divided into morphemes, which are basic units, and information such as a part-of-speech and conjugation system is added.
Then, in step S30, the sentence morphologically analyzed in step S20 is received, the syntactic structure of the input sentence is analyzed from the information of the result of morphological analysis, and the information is passed to the next step S40. In step S40, the semantic relationship of the input sentence is analyzed from the analysis result of the syntax analysis process (step S30).

In step S50, in the process of analyzing the input sentence by the syntax analysis process (step S30) and the semantic analysis process (step S40), if the input sentence contains omitted information, the information is complemented. To do. Here, the optional information holding unit (stored in the dictionary / knowledge base) prepared for efficiently restoring the information omitted in the input sentence is referred to. The abbreviated character information storage unit stores abbreviated character strings and their corresponding regular character strings in a table format, and if an abbreviated character string is searched for, it becomes possible to extract the corresponding regular character string. There is.

In step S60, a semantic analysis process is performed using the information from the abbreviation complementation part (step S40).
The supplementary information and the contents of the ambiguous portion temporarily determined by the device are added to the original input sentence and displayed from the result of the semantic analysis obtained from. Then, the control proceeds to the next Step S70. Here, information that is insufficient for searching the database 8 or ambiguous parts in the input sentence is determined, and if necessary, the user is asked again, and the input sentence receiving unit 1
Return to 0. When it is determined that the information is sufficient for searching the database 8, the process proceeds to the next step S80,
The database 8 is searched according to the search condition determined by the question content determination process (step S70). When the search is completed, the process proceeds to step S90, the result is displayed on the CRT 5, and the information displayed in the dialogue history holding unit (secured in the RAM 2) is delivered. Then, the process returns to step S10 to end the information retrieval or to continue repeating the question and answer. The dictionary / knowledge base 7 is used in the sentence analysis process including steps S20 to S40.

Next, the dialogue in this embodiment will be described by comparing the dialogue example of FIG.

U5 question sentence (U5) in FIG. 2 What is the light around 6 o'clock? Is input to the device.

At this time, in the embodiment, the input portion reception, the morphological analysis, the syntactic analysis, and the semantic analysis of steps S10 to S40 are performed, and the semantic representation of FIG. 2 is obtained.

According to this result, the time and type in the train information are obtained, but there is no other information. Therefore, assuming that there is an abbreviation, control is passed to the abbreviation complementation processing (step S50). In the abbreviation complementation process, the knowledge of the question about the train information is retrieved from the dictionary / knowledge base 7 into an abbreviated information holding unit (secured in the RAM 3).
FIG. 3 shows knowledge of train information and omissible information (in this case, an operation of connecting time and other train information). By comparing this information with the semantic representation of FIG. 2, the complemented abbreviation of FIG. 4 is obtained. Here, it is supplemented that the operation relationship connecting time and type is (departure time) or (arrival time) operation. Here, the device tentatively interprets that it is the departure time. Then, the supplemented meaning expression is passed to the sentence analysis result display processing (step S60). In a sentence analysis result display processing, based on the semantic representation, as a result of analysis of the question sentence of U1 the (S5 ') about 6 o'clock in the <starting> Hikari
<I want to search> ? Is displayed on the CRT 5. In this display, the part enclosed in parentheses <> is the complemented result of the information required by the device for searching, and the underlined part is the expression added to the original input sentence. By viewing this display, the user can easily recognize the information complemented by the device. Thereafter, the control is passed to the question content determination process (step S70). Here, the user is confirmed for (U5 '). If necessary, the supplementary result is corrected by the question response, and after the correction is completed, the result is passed to the information retrieval process (step S80). In this information retrieval process, (time: departure time: 6:00), (type:
A search is performed under the condition of "Hikari" and the search result is passed to the result display process (step S90). In the result display process, there are two messages (S8) that are the search results. Is displayed. Then, the result is passed to the dialogue history holding unit, and control is performed to receive an input sentence (step S
Return to 10).

The dialog when the above-mentioned processing of the present apparatus is performed is as follows.

(U5) What about the light around 6 o'clock? (S5 ') the <starting> Hikari to about 6 o'clock
<I want to search> ? (U6) Yes (S8) There are two.

Thus, in the apparatus of this embodiment,
Conventionally, six dialogues were required from U5 to S8, whereas four dialogues are completed.

[Description of Other Embodiments] In the above embodiment, the sentence analysis result display section displays the device complement result on the next line as in the message S5 '. However, the present invention is not limited to this. Various display methods are possible. For example, (U5)
It is also possible to give the effect that the information complemented by the device is purchased and displayed in the question text by writing (S5 ') over. According to this, the result of the above dialogue is as follows.

(S5 ') About 6 o'clock <departure>
The Hikari <want to find>? (U6) Yes (S8) There are two.

In this case, the dialogue is completed three times.

In the sentence analysis result display process, the parentheses <> and the underline are used to indicate the abbreviated expression, but the present invention is not limited to this, and various expression methods are possible. For example, black-and-white inverted characters may be used instead of parentheses, and in a display device capable of color display, it is also possible to change the display color.

Furthermore, taking the input sentence U5 as an example, it is also possible to perform omission completion based on the information of the dialogue history holding unit, instead of omission complement based on the train information stored in the dictionary / knowledge base 7. For example, as shown in FIG. 5, after the search proceeds from message S5 to message S8, input sentence U9
Explain when the question is input.

Through the morphological analysis process (step S20), the syntactic analysis process (step S30), and the semantic analysis process (step S40), a semantic analysis result as shown in FIG. 5 is obtained. After that, by the abbreviation complementing process (step S50), using the same mechanism as the mechanism described in the first embodiment, among the results shown in FIG. 7, (S9 ′) <Buffet> of Imperial Hotel Hikari
No.> is complemented by the underlined part of <I want to search> .

At the same time, the abbreviation complementing process (step S50) uses the information stored in the dialogue history holding unit to
The dialogue history as shown in FIG. 6 is taken out to hold the optional information. With this information, there is a possibility that the search result of the message S8 may be further searched depending on the past search conditions up to the message S8. In the abbreviation complementation process, it is determined that the result of the message S8 is omitted as a restriction condition, and the semantic analysis result as shown in FIG. 7 is passed to the sentence analysis result display process (step S60). The sentence analysis result display process, on the basis of this meaning representation, as an analysis result of the query sentence input sentence U9, (S9 ') in <two S8>, of Imperial <bi
Display <Hikari-go> of <uffe><Search> . Thereafter, the control is passed to the question content determination process (step S70). Here, the user is asked to confirm the message S9 '. If necessary, the supplementary result is corrected by a question answer, and after the correction is completed, the result is passed to the information retrieval process (step S80). In this information search processing, a search is performed under the conditions of (restriction condition: S8) and (attribute: buffet: Imperial Hotel), and the search result is passed to the result display processing (step S90). Here is the result of that search, (S12) 6:00 AM Hikari No. 16 from Tokyo Station. Is displayed. Then, the result is passed to the dialogue history holding unit, and control is performed to receive an input sentence (step S1).
Return to 0).

As described above, the dialogue when using this device is as follows.

(S8) There are two.

(S9 ') <Hikari-go> of <Buffet> of < Imperial Hotel > was <searched in < 2 of S8 >.
I> (S12) 6:00 AM Hikari No. 16 departing from Tokyo Station.

Thus, in FIG. 7, U9 through S1
Up to 2, 6 dialogues have been completed in 3 dialogues.

As described above, according to this embodiment,
By displaying the information supplemented by the abbreviation complementation processing in the sentence analysis result display processing, the following effects can be obtained.

1. The user can easily confirm the information complemented by the device.

2. By displaying it in addition to the input sentence, it becomes possible to eliminate unnecessary communication in the confirmation work performed between the device and the user.

3. Since the display format can be changed, the user can distinguish and recognize various complementary processes of the device.

Further, in the embodiment, the device for handling the train schedule is provided, but of course the present invention is not limited to this, and it may be realized by incorporating a program.

[0048]

As described above, according to the present invention, when a character string of abbreviated name is included in a sentence input for searching, it is possible to easily confirm the original intention of the input sentence. Become.

[Brief description of drawings]

FIG. 1 is a flowchart showing a processing flow of a natural language processing apparatus according to this embodiment.

FIG. 2 is a diagram for explaining operation contents in the present embodiment.

FIG. 3 is a diagram for explaining operation contents in the present embodiment.

FIG. 4 is a diagram for explaining operation contents in the present embodiment.

FIG. 5 is a diagram for explaining operation contents in another embodiment.

FIG. 6 is a diagram for explaining the operation content in another embodiment.

FIG. 7 is a diagram for explaining the operation content in another embodiment.

FIG. 8 is a block configuration diagram of an apparatus according to an embodiment.

FIG. 9 is a flowchart showing a processing flow of a conventional natural language processing device.

FIG. 10 is a diagram showing an example of a processing operation of a conventional natural language processing device.

[Explanation of symbols]

 1 CPU 2 ROM 3 RAM 4 keyboard 5 CRT 6 VRAM 7 dictionary / knowledge base 8 database

Claims (1)

[Claims] 1. A natural language processing device for retrieving information by a sentence input in a natural language, a storage unit for storing a simplified name and a canonical name in association with each other, and an input unit for inputting a sentence for retrieval. A determining unit that determines whether or not there is a character string of the abbreviated name stored in the storage unit in the character string forming the input sentence; and when the determining unit determines that there is a character string of the abbreviated name,
A natural language processing apparatus comprising: a complementing unit that complements the regular character string stored in the storage unit with respect to the character string, and a display unit that displays a sentence complemented by the complementing unit.