JPH08180064A - Document retrieval method and document filing device - Google Patents

Abstract

PURPOSE: To highly accurately retrieve picture data provided with a retrieval key word by developing an inputted retrieval key word to the plural retrieval key words by using an erroneous recognition table and retrieving character- recognized and stored document data based on the developed plural key words. CONSTITUTION: Area separation is performed to input pictures, a text area is obtained, character recognition is performed to the text area and characters and character strings for which erroneous recognition occurs at the ratio of equal to or more than a prescribed threshold value for the character and character string under consideration in the character recognition are stored in the erroneous recognition table 107a corresponding to the character and character string under consideration. Then, the retrieval key word inputted by a scanner 101 and a mouse 103, etc., is developed to the plural retrieval key words by using the erroneous recognition table 107a, and based on the developed plural key words, the character-recognized and stored document data are retrieved and displayed on a display 108. Thus, the picture data of the retrieval key word are highly accurately retrieved.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document retrieval method and a document filing apparatus, and more particularly, to a document retrieval method and a document filing capable of inputting and accumulating a document image and retrieving a character string in the accumulated image. It relates to the device.

[0002]

2. Description of the Related Art A conventional document filtering apparatus having a character string search function is obtained by capturing printed characters with a scanner and accumulating them as image data, and recognizing a text area in the image data. A character code is stored as text data, and when a character string is searched, a search character string (search keyword) is collated with the stored text data to search for a character string in an image.

FIG. 9 is a flowchart generally showing the flow of processing when a search is performed. The conventional search process will be described below with reference to the flowchart of FIG.
In step S901, a character string as a search keyword is input. In step S902, "1" indicating the first data is set to the counter i in order to instruct the order of the search data to be searched from now on. In step S903, the text data part of the stored i-th data is compared with the input search keyword. This matching performs a so-called full-text search to check whether or not the search keyword is included in the text data portion as it is. In step S904, it is determined whether the text data includes the search keyword.

If the search keyword is included, the process proceeds to step S905. In step S905, all the image data including the found text data is decompressed and displayed on the display, and the process proceeds to the next step S906. Here, the case where the search keyword is included in the text data portion will be referred to as “hit”. If the search keyword is not included in the text data portion in step S904, the process proceeds to step S906. In step S906, it is determined whether there is more data. If there is no more data, it ends. If there is data, the counter i is incremented by one in step S907, and then step S9.
Return to 03.

As described above, it is possible to find an image of data containing the search keyword as a character image and display it on the display.

[0006]

However, in the above conventional example, since the search keyword is collated with the text data as it is, when the text data contains an erroneous character due to an erroneous recognition in character recognition, There was a drawback that the collation did not hit and the correct data was not retrieved.

The present invention eliminates the above-mentioned conventional drawbacks and a document search method capable of finding image data containing a search keyword with high accuracy even if character recognition is not 100% correct at the time of registration of document image data. A document filing device is provided.

[0008]

In order to solve the above-mentioned problems, the document search method of the present invention is characterized in that a character recognition causes a character and a character string to be erroneously recognized at a ratio of a predetermined threshold value or more with respect to a target character and a character string. A character string is stored as an erroneous recognition table corresponding to the target character and the character string, the input search keyword is expanded into a plurality of search keywords using the erroneous recognition table, and the expanded plurality of keywords are Based on this, character recognition is performed to retrieve stored document data.

Here, the document data includes text data and image data, the text data is searched, and the document data including the searched text data is further output. Further, in the erroneous recognition table, a predetermined number of characters and character strings corresponding to the target character and character string are arranged in descending order of probability of error. Further, the plurality of expanded search keywords are limited to a predetermined number.

Further, the document filtering device of the present invention is a document filtering device for character-recognizing and filing a text portion of a document image, and erroneously recognizing a target character and a character string by a character recognition at a ratio of a predetermined threshold value or more. A character and a character string in which the occurrence of the error occurs is stored in association with the target character and the character string, and a plurality of input search keywords are input based on the misrecognized character and the character string stored in the storage means. It is characterized in that it is provided with a expanding means for expanding into a search keyword and a searching means for recognizing the character data and searching the stored document data based on the expanded plurality of keywords.

Here, the document data includes text data and image data, and the search means further comprises output means for searching the text data and outputting the document data including the searched text data. Further, the storage means is an erroneous recognition table arranged in an order in which a predetermined number of characters and character strings corresponding to the target character and character string are in a high probability of being mistaken. Further, the plurality of search keywords expanded by the expanding means are limited to a predetermined number.

[0012]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. <Configuration Example of Document Filing Device> FIG. 1 is a block diagram showing the system configuration of a document filing device for explaining an embodiment.

In FIG. 1, 101 is a scanner which irradiates an image original with light and reads the reflected light to convert it into an electric signal, and 102 converts the electric signal obtained by the scanner 101 into a binary digital electric signal, A scanner interface circuit for transmitting to other device components, 103 is a pointing device (mouse, etc.) for inputting desired coordinates on the window of the display, 104 is a signal from the pointing device 103, and receives it. A pointing device interface (I / F) circuit for transmitting to the device component of the device, 105 is a CP for controlling the entire device and executing character cutout processing and recognition processing.
U and 106 are ROMs that store control programs executed by the CPU 105, various processing programs and font data, and 107 is a RAM used as a work area for developing character images and character recognition processing. The error recognition table 107a of the embodiment is included. Also,
Reference numeral 108 is a display for displaying an input image and a recognition result, and 109 is a display interface circuit. The display 108 displays the image of the VRAM area stored in the predetermined address area of the RAM 107. Reference numeral 110 is an external storage device such as a hard disk in which registered data is stored, and 111 is its interface. Reference numeral 112 is a bus for connecting each device constituent element.

<Error Recognition Table> Next, the error recognition table 107a (referred to as a confusion matrix) of this embodiment will be described. The erroneous recognition table in the present invention is a table in which characters that are easily mistaken by character recognition are registered in advance. FIG. 3 shows an example of the misrecognition table (confusion matrix). In the misrecognition table 107a, as a result of a recognition experiment for a large number of learning data, characters in which misrecognition occurs at a rate of a certain threshold value or more are registered.

The configuration will be described with reference to FIG. The vertical line from the reference numeral 301 is the target character, and the reference numeral 30
Characters that are easily mistaken from 2 are arranged in descending order from the one with the highest probability of being mistaken. The number of characters that are easy to make a mistake is a fixed value of N at maximum, and if there are no N characters that are easy to make a mistake, leave blank. Further, when there is no character that is erroneously recognized when the value is equal to or larger than the threshold value, the erroneously recognized character column is blank as indicated by reference numeral 303.

The erroneous recognition table (confusion matrix) described above is prepared in the ROM 106 or the RAM 107 or the external storage device 110. In the above example, R
It is assumed to be in AM107. <Example of Operation of Document Filing Device> (Registration Process) FIG. 6 is a flowchart showing the flow of a process at the time of registering a document in a document filing device having a search function for a character string in an image. Hereinafter, FIG. 6 will be described.

In step S601, a document to be registered is read by a scanner and converted into binary image data. In the next step S602, the obtained input image is subjected to area separation, and only the text area in which the character image exists is extracted. Extracting only the text region is possible by extracting the connected component of the black pixels in the binary image data and combining only those that are estimated to be characters, which is an existing technique.

In the next step S603, character recognition processing is performed on the text area to obtain text data. This text data is used for matching with a search keyword at the time of search. In step S604, the entire image data is compressed. In this embodiment, it is assumed that MMR, which is optimal for binary images, is used. In step S605, the compressed image data and text data are integrated, and a header describing the date, registrant name, data size, etc. is added and saved in the external storage device.

FIG. 7 schematically shows how the regions are separated. Area separation is performed on the input image 701 to obtain a text area 702. This text area 70
Character recognition is performed on 2 to obtain text data. On the other hand, the compressed original image data is paired with the text data, and a header is further added to add one data 7
Accumulate as 03.

FIG. 8 shows the stored data. The first data is stored in 801 and the second and subsequent data are stored in order from 802. (Search Processing) FIG. 2 is a flowchart showing a processing procedure for searching when an erroneous recognition table is used. Less than,
A search using the erroneous recognition table will be described with reference to the flowchart of FIG. In step S201,
For example, assume that the character string “cup” is entered as a search keyword.

In the next step S202, the erroneous recognition table 107 is sequentially used with the constituent characters of the search keyword as a key.
Search for a (confusion matrix), and in "cup" there is a similar character "power" (kanji) for "ka" (katakana) and similar character for "tsu" Without it, it can be seen that there is a similar character “B” for “P”. Therefore, as the search keyword expanded from the input search keyword “cup”, there are four combinations of “cup”, “power up”, “cobb”, and “power web” that are combinations including all similar characters. The search keyword is generated as shown in FIG.

In step S203, step S in FIG.
Similarly to 902, a counter i indicating the order of search data is
The first data (data 801 in the example of FIG. 8) is set to "1". In the next step S204, the four search keywords expanded in step S202 (hereinafter referred to as expanded keywords) are used to collate with the text data. The matching method is full-text search, and full-text search is performed as in step S903 in FIG. Step S20
At 5, it is determined whether the text data includes the expansion keyword.

If the expansion keyword is included, the process proceeds to step S206. In step S206, all image data including the found text data is decompressed and displayed on the display, and the process proceeds to the next step S207. Here, the case where the expansion keyword is included in the text data portion will be referred to as “hit”. When the expansion keyword is not included in the text data part in step S205, the process proceeds to step S207. In step S207, it is determined whether there is more data. If there is no more data, it ends. If there is data, the counter i is incremented by 1 in step S208, and then step S2
Return to 04.

As described above, it is possible to find the image of the data in which the expansion keyword is included as the character image and display it on the display. As described above, by using the misrecognition table 107a (confusion matrix) and using the expansion keyword expanded to the similar character, the following advantages occur. That is, as shown in FIG. 5, for example, even if a part of the input text data, "cup", is mistakenly recognized as "power" and registered, the search keyword "cup" is used when searching. When the search is performed, the search keyword is expanded into the expanded keyword by the misrecognition table, and the target text can be searched. That is, even if the recognized and input text has an error, the target text data can be correctly hit and searched.

In the above-described embodiment, the search keyword is unconditionally expanded using the misrecognition table.
All the expansion keywords considered as a combination of the results are generated. However, if there are many characters that can be expanded, the number of possible combinations will greatly increase, so expansion keywords may be generated under the rule that K or more replacements are not performed.

In fact, since character recognition is not so often mistaken, it can be assumed that there are K or less mistakes in one word that can be a search keyword. As K, for example, the number of characters of the search keyword may be L, and may be a number determined as follows. K = L / 4 Further, in this embodiment, the misrecognition table corresponding to each character is prepared, but the misrecognition table corresponding to a plurality of characters or character strings may be prepared.

Furthermore, the present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0028]

According to the present invention, even if the character recognition is not 100% correct when registering the document image data, the character string of the search keyword is expanded according to the erroneous recognition table and then collated, so that the search is performed with higher accuracy. There is an effect that image data including a keyword can be searched for.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of a system for explaining a document filing device according to an embodiment.

FIG. 2 is a flowchart of a search according to this embodiment.

FIG. 3 is a diagram showing an example of an erroneous recognition table in the present embodiment.

FIG. 4 is a diagram showing an example of keyword expansion at the time of search according to this embodiment.

FIG. 5 is a diagram showing an example of text data of the present embodiment.

FIG. 6 is a flowchart showing a flow of processing at the time of document registration in the character filing device.

FIG. 7 is a diagram schematically showing area division in registration.

FIG. 8 is a diagram showing a structure of data registered and stored.

FIG. 9 is a flowchart showing a flow of processing when performing a conventional search.

[Explanation of symbols]

 101 Scanner 102 Scanner Interface Circuit 103 Pointing Device 104 Pointing Device Interface Circuit 105 CPU 106 ROM 107 RAM 107a False Recognition Table 108 Display 109 Display Interface Circuit 110 External Storage Device 111 External Storage Device Interface Circuit

Claims (8)

[Claims] 1. A character and a character string, which are erroneously recognized by the character recognition with respect to the target character and the character string at a ratio of a predetermined threshold value or more, are stored as an erroneous recognition table corresponding to the target character and the character string. A document characterized in that the input search keyword is expanded into a plurality of search keywords by using the misrecognition table, and the stored document data is searched by character recognition based on the expanded plurality of keywords. retrieval method. 2. The document data includes text data and image data, the text data is searched, and the document data including the searched text data is further output. Document search method. 3. The document according to claim 1, wherein in the erroneous recognition table, a predetermined number of characters and character strings corresponding to the target character and character string are arranged in descending order of probability of error. retrieval method. 4. The document search method according to claim 1, wherein the plurality of expanded search keywords are limited to a predetermined number. 5. A document filing apparatus for character-recognizing and filing a text portion of a document image, wherein a character and a character string, which are erroneously recognized by the character recognition with respect to a target character and a character string, at a ratio of a predetermined threshold value or more, Storage means for storing the noted character and character string, and expanding means for expanding the input search keyword into a plurality of search keywords based on the misrecognized character and character string stored in the storage means. A document filtering device, comprising: a search unit that performs character recognition based on the expanded plurality of keywords and searches the stored document data. 6. The document data includes text data and image data, and the search means further comprises output means for searching the text data and outputting document data containing the searched text data. The document filtering device according to claim 5, wherein 7. The erroneous recognition table, wherein the storage means is arranged in an order in which a predetermined number of characters and character strings corresponding to the target character and character string are arranged in descending order of probability of error. Document document described device. 8. The document filtering apparatus according to claim 5, wherein the plurality of search keywords expanded by the expanding means are limited to a predetermined number.