JP2011065597A - Device and data searching, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain a desired search result with less search omission even when an error recognition character is included in text data. <P>SOLUTION: A memory 16 includes an error recognition database 23 where characters with high possibility for error recognition are prestored as error recognition characters concerning the respective characters. When a search word is input through an input part 13, a CPU 11 exchanges the respective characters of the search word with the error recognition characters stored in the database 23, thereby to search a text by adding the search word after character exchange to the search words. Consequently, the desired search result with less search omission is obtained even when the error recognition character is included in the text data. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a data search apparatus, a data search method, and a program that are used in an information processing apparatus such as a PC (Personal Computer), for example, and perform data search on the result of character recognition.

  Conventionally, there are the following Patent Documents 1 and 2 as techniques for performing a search on data obtained by character recognition of handwritten characters.

  In Patent Document 1, when a search target includes not only text but also a character image (image), both the text and the character image are obtained by making a search word into an image and performing a search by pattern matching. Is made searchable.

  Patent Document 2 discloses that search word input means is character recognition, and a plurality of recognition results are prepared and searched to reduce search omissions.

JP-A-8-55134 JP 2006-163830 A

  Usually, when a document in which handwritten characters or printed characters are described is recognized, not all characters are recognized correctly, and some characters are often erroneously recognized. If text data including such misrecognized characters is used as a search target, a search omission may occur even if a search word is correctly input. In Patent Documents 1 and 2 described above, there is no particular mention of processing for such erroneously recognized characters.

  The present invention has been made in view of the above points, and a data search apparatus capable of obtaining a desired search result with reduced search omission even when erroneous recognition characters are included in text data. An object of the present invention is to provide a data search method and program.

  The data search device according to the present invention includes a search target acquisition unit that acquires text data that has been character-recognized as a search target, and an error in which characters that are likely to be erroneously recognized are stored as erroneously recognized characters in advance. A recognition database, an input means for inputting a search word, a character replacement means for replacing each character of the search word input by the input means with an erroneous recognition character stored in the erroneous recognition database, and the search word And a search processing means for searching for the text data by adding a search term after the character replacement obtained by the character replacement means.

  In addition, the data search method according to the present invention acquires character-recognized text data as a search target, and a database for erroneous recognition in which characters that are likely to be erroneously recognized are stored as erroneously recognized characters in advance for each character. To replace each character of the input search word with a misrecognized character stored in the database for misrecognition, and to search the text data by adding the search word after the character replacement to the search word. Features.

  The program according to the present invention is a data search program executed by a computer, and the computer has a function of acquiring text data that has been recognized as a search target, and each character has been erroneously recognized in advance. A function of replacing each character of the input search word with a misrecognized character stored in the misrecognition database, using a misrecognition database in which characters that are likely to be stored as misrecognized characters, and the search And a function for searching the text data by adding a search word after the character replacement to a word.

  According to the present invention, data search is performed by replacing each character of the input search word with a character that is highly likely to be erroneously recognized. Thereby, even if a misrecognized character is included in the text data, it is possible to reduce a search omission and obtain a desired search result.

FIG. 1 is a block diagram showing a configuration of a data search apparatus according to the first embodiment of the present invention. FIG. 2 is a diagram showing a configuration of a search target database provided in the data search apparatus according to the embodiment. FIG. 3 is a diagram showing a configuration of a misrecognition probability database provided in the data search apparatus according to the embodiment. FIG. 4 is a diagram illustrating a functional configuration of a CPU provided in the data search apparatus according to the embodiment. FIG. 5 is a flowchart showing the processing operation of the data search apparatus in the embodiment. FIG. 6 is a diagram showing an example of a search word in the same embodiment. FIG. 7 is a diagram illustrating an example of a misrecognized character corresponding to each character of a search word in the embodiment. FIG. 8 is a diagram showing an example of a misrecognized character corresponding to each character of the search word in the embodiment. FIG. 9 is a flowchart showing the processing operation of the data search apparatus according to the second embodiment of the present invention. FIG. 10 is a diagram showing an example of search candidates in the embodiment.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(First embodiment)
FIG. 1 is a block diagram showing a configuration of a data search apparatus according to the first embodiment of the present invention.

  The data search apparatus according to the present embodiment includes a CPU 11, a display unit 12, an input unit 13, a scanner unit 14, a storage device 15, a memory 16, and a medium reading device connected to the CPU 11 via a system bus 10. Device 17.

  The CPU 11 controls the entire apparatus, and executes a series of processes related to data retrieval by starting a program according to an input instruction. The display unit 12 includes a display device such as a CRT (Cathode-ray tube) or an LCD (Liquid Crystal Display), and displays various data including a data search result. The input unit 13 includes an input device such as a keyboard and a mouse, and inputs and instructs various data including a search term. The scanner unit 14 is an apparatus for capturing and capturing documents and photographs.

  The storage device 15 includes, for example, a flash memory, a hard disk device, and the like. The storage device 15 includes an image database (image DB) 21, a search target database (search target DB) 22, and an error recognition probability database (error recognition probability DB) 23.

  The image database 21 stores various image data. The search target database 22 stores data used as a search target. In the present embodiment, text data obtained as a result of reading a newspaper page by the scanner unit 14 and character recognition of an article on the page is stored in the search target database 22 as a search target. Hereinafter, “text data obtained as a result of character recognition” is simply referred to as “text data with character recognition”. It is assumed that the text data includes misrecognized characters.

  Further, the image related to the article is captured by the scanner unit 14 and stored in the image database 21. At that time, as shown in FIG. 2, the search target database 22 stores link information indicating the storage destination of the image data together with the text data that has been character-recognized.

  The misrecognition probability database 23 stores the probability that the character recognition algorithm used when the search target database 22 was created will misrecognize. Specifically, for various characters such as hiragana, katakana, kanji, and alphabet, a table is stored in which characters that are erroneously recognized and their probabilities are associated with each other. This table is separately generated from the corpus data of the character image. However, the generation method is not particularly limited, and any method may be used.

FIG. 3 shows an example of the erroneous recognition probability database 23 for hiragana characters.
Hiragana characters “A”, “I”, “U”,... Are associated with characters that are misrecognized and their probabilities. For example, if the hiragana character is “a”, the probability of being erroneously recognized as a character such as “o” is (20), and the probability of being erroneously recognized as a character such as “me” is (7). The probability of erroneous recognition is (3).

  Here, as shown in FIG. 4, the CPU 11 as the main control device has a character replacement unit 11a, a search processing unit 11b, as a functional configuration when text data including such misrecognized characters is a search target. A search candidate generation unit 11c is provided.

  The character replacement unit 11 a performs a process for replacing the character of the search term input by the input unit 13 with a misrecognized character stored in the search target database 22. The search processing unit 11b performs processing for searching for text data by adding the search term after the character replacement obtained by the character replacement unit 11a to the input search term. The search candidate generation unit 11c performs processing for generating a predetermined number of search candidates at the time of batch search in the second embodiment to be described later.

  In FIG. 1, the memory 16 includes a ROM, a RAM, and the like, and stores various data necessary for the processing operation of the CPU 11. The memory 16 includes a program storage unit 24, an input buffer 25, an output buffer 26, a search number storage unit 27, and a search candidate storage unit 28.

  The program storage unit 24 stores various programs such as an application program for character recognition in addition to a program for realizing the present invention. The input buffer 25 stores a search term input from the input unit 13. The output buffer 26 stores data obtained as a search result.

  The number-of-search storage unit 27 stores the number of searches designated by the user (number extracted from the text data as a search result). Further, the search candidate storage unit 28 stores data of a predetermined number of search candidates generated at the time of batch search in the second embodiment to be described later.

  The medium reading device 17 is a device for reading various data including a program recorded on the recording medium 18. Examples of the recording medium 18 include a magnetic disk (flexible disk, hard disk, etc.), an optical disk (CD-ROM, DVD-ROM, etc.), and a semiconductor memory.

Next, the operation of this apparatus will be described.
FIG. 5 is a flowchart showing the processing operation of the data search apparatus in the first embodiment. The processing shown in this flowchart is recorded on the recording medium 18 in the form of a program readable by the CPU 11 and installed through the medium reading device 17 or stored in advance in the program storage unit 24 of the memory 16. Yes.

  It is assumed that text data obtained by reading a large number of newspaper pages with the scanner unit 14 and character-recognizing is stored in the search target database 22 for each article, and a desired article is searched from the text data. Note that the text data includes misrecognized characters in terms related to the article.

  First, as an initial setting, the user inputs a term related to a desired article as a search term through the input unit 13 (step S11), and designates the number of articles to be extracted as the number of searches (step S12). As a result, the CPU 11 stores the search word input from the input unit 13 in the input buffer 25 of the memory 16 and stores the designated search number in the search number storage unit 27 of the memory 16, and then searches as follows. Execute the process.

  That is, the CPU 11 searches the search target database 22 based on the search word stored in the input buffer 25, and searches for article data that matches the search word from the text data stored in the search target database 22. And stored in a work area (not shown) of the memory 16 (step S13). In this case, all relevant article data is extracted from the text data, and the number of the data is counted as the number of searches this time.

  Here, if the current number of searches does not reach the number of searches (for example, 100) stored in the search number storage unit 27 (No in step S14), the CPU 11 determines each character constituting the input search word. Is replaced with the corresponding misrecognized character in the misrecognition probability database 23 to generate a search word with a high probability of misrecognition (step S15), and text data is generated using the generated new search word. Search again (step S13).

  A process for generating a search word with a high probability of erroneous recognition will be specifically described with reference to FIGS.

  If a character string such as “election violation” is input as a search word as shown in FIG. 6, the characters “selection”, “lift”, “fault”, “anti” that constitute this “election violation” As a result, erroneous recognition character data as shown in FIG. 7 is obtained.

  The numbers in parentheses in the figure represent the probability of misrecognition. For example, there are misrecognized characters such as “achievement”, “arrest”, and “speed” for “Select”, each of which is misrecognized. It is shown that the probability of being (20), (12), (5).

  Here, the character with the highest misrecognition rate among the misrecognized characters corresponding to each character of the search term “election violation” is “sent” having a probability (23). This “sent” is a misrecognized character corresponding to “wrong” of the search term “election violation”. Therefore, as shown in FIG. 8, a new search term such as “election sentiment” is generated by replacing “fault” in “election violation” with “send” (step S15), and text data is generated using the search term. Is searched again (step S13).

  If the specified number of search results are not obtained by this re-search (No in step S14), “luck” with the next highest probability of misrecognition is selected, and “wrong” is replaced with “luck”. Then, a new search word such as “election luck” is generated (step S15), and the text data is re-searched with the newly generated search word (step S13). If the specified number of search results is not obtained even after this re-search, select the next “Dachi” with the highest probability of misrecognition, replace “Select” with “Dai”, The text data is re-searched with a new search term such as "".

  If a specified number of search results cannot be obtained by replacing one character at a time, a new search word is generated by appropriately combining the characters in order from the character with the highest probability of misrecognition. . In this case, the combination is arbitrary, but first, the character is searched for in combination with two characters in descending order of the probability of misrecognition. It is preferable to perform the search again by increasing the number of characters to be replaced. This is because, in general, all characters of a term are rarely misrecognized, and only some characters are often misrecognized.

  In this way, the text data is searched again by exchanging characters in descending order of the probability of erroneous recognition. When the specified number of search results are obtained, specifically, when the total value of the search numbers obtained at the current time reaches the search number (for example, 100) stored in the search number storage unit 27 ( In step S14, the CPU 11 reads these search results from the work area of the memory 16, develops them in the output buffer 26, and displays them in a predetermined format on the display unit 12 so that the user can confirm them. Output (step S16).

  Here, a predetermined number (100) of articles related to “election violation” is extracted from the text data of the search target database 22 and displayed on the screen of the display unit 12. If an image is included in the extracted article, the corresponding image data is read from the image database 21 based on the link information added to the text data of the article and displayed together with the text data. It will be.

  As described above, according to the present embodiment, when searching for data that matches the search word from text data that has been character-recognized, it is not replaced with a simple character, but replaced with a character that is likely to be erroneously recognized. By continuing the data search, even if a misrecognized character is included in the text data, it is possible to reduce a search omission and obtain a desired search result.

  In addition, for each character, the misrecognized character and the probability of misrecognizing the character are stored in the misrecognition database (the misrecognition probability database 23) in advance, and when replacing each character in the search term, By referring to the recognition database and sequentially replacing characters with a high probability of erroneous recognition, efficient search can be performed on text data including erroneously recognized characters.

  Also, by continuing the search by replacing the characters until a certain number of search results are obtained, even if they are misrecognized as different characters at each location, search without leaking them Is possible.

(Second Embodiment)
Next, a second embodiment of the present invention will be described.

  In the first embodiment, the text search is performed while sequentially replacing each character of the search word until a certain number of search results are obtained. However, in the second embodiment, a plurality of search candidates from the search word in advance. Is generated, and a batch search is performed using these search candidates.

  Since the apparatus configuration is the same as that of the first embodiment, here, the search processing as the second embodiment will be described with reference to FIGS.

  FIG. 9 is a flowchart showing the processing operation of the data search apparatus in the second embodiment. The processing shown in this flowchart is recorded in the recording medium 18 in the form of a program readable by the CPU 11 and installed through the medium reading device 17 or stored in advance in the program storage unit 24 of the memory 16. Yes.

  Assume that text data obtained by character recognition of a newspaper page is stored in the search target database 22, and a desired article is searched from the text data. Note that the text data includes misrecognized characters in terms related to the article.

  First, as an initial setting, a user inputs a term related to a desired article as a search word through the input unit 13 (step S21). Thus, the CPU 11 stores the search word input from the input unit 13 in the input buffer 25 of the memory 16 and then executes the following search process.

  That is, the CPU 11 generates a predetermined number of search candidates from the search terms stored in the input buffer 25 (step S22). Specifically, a predetermined number of search candidates are generated by replacing each character constituting the search word with the corresponding erroneously recognized character in the search target database 22 and stored in the search candidate storage unit 28 together with the input search word. Remember.

  FIG. 10 shows an example of search candidates stored in the search candidate storage unit 28. The numbers in parentheses indicate the priority order of search candidates.

  For example, a search term such as “election violation” is input, and the characters “selection”, “lift”, “difference”, and “anti” constituting this “election violation” are checked in the erroneous recognition probability database 23, Assume that an erroneously recognized character as shown in FIG. 7 is obtained. Among these, “sent”, “luck”, “achi”, “spicy”, which have a high probability of erroneous recognition, are replaced with corresponding characters of the search word to generate a search candidate like the example of FIG.

  Basically, it replaces the characters with high probability among the misrecognized characters with priority, but the method of combining the misrecognized characters is arbitrary, and the number of candidates can also be set arbitrarily. is there.

  When a predetermined number of search candidates are generated in this way, the CPU 11 collectively searches the search target database 22 based on the original search words and each search candidate stored in the search candidate storage unit 28, and the search target database. Data that matches either the search word or each search candidate is extracted from the text data stored in 22 as a search result, and stored in a work area (not shown) of the memory 16, for example (step S23).

  That is, in the example of FIG. 10, the original search word “election violation”, or any of the conversion candidates “election refusal”, “election refusal”, “advance violation”, “admission refusal” Matching data is extracted as a search result.

  The CPU 11 reads out these search results from the work area of the memory 16, expands them in the output buffer 26, displays them in a predetermined format on the display unit 12, and outputs them so that they can be confirmed by the user (step S 16). .

  In this way, even when a predetermined number of search candidates are generated from the search words in advance and a text search is performed using them, the text data to be searched is erroneously recognized as in the first embodiment. Even when characters are included, it is possible to reduce a search omission and obtain a desired search result.

  In each of the above embodiments, a case has been described in which a newspaper page is captured by a scanner and text data obtained by character recognition is used as a search target, but the present invention is not limited to a newspaper page. The present invention can be similarly applied to text data obtained by character recognition of some document. Alternatively, text data that has been character-recognized in advance may be acquired from the outside via a recording medium or a communication medium.

  In short, the present invention is not limited to the above-described embodiments as they are, and can be embodied by modifying constituent elements without departing from the scope of the invention in the implementation stage. Various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the respective embodiments. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

  In addition, the method described in the above-described embodiment is a program that can be executed by a computer, such as a magnetic disk (flexible disk, hard disk, etc.), an optical disk (CD-ROM, DVD-ROM, etc.), a semiconductor memory, etc. The program can be written on a medium and applied to various apparatuses, or the program itself can be transmitted through a transmission medium such as a network and applied to various apparatuses. A computer that implements this apparatus reads a program recorded on a recording medium or a program provided via a transmission medium, and performs the above-described processing by controlling operations by this program.

  DESCRIPTION OF SYMBOLS 10 ... System bus, 11 ... CPU, 11a ... Character substitution part, 11b ... Search processing part, 11c ... Search candidate production | generation part, 12 ... Display part, 13 ... Input part, 14 ... Scanner part, 15 ... Memory | storage device, 16 ... Memory: 17 ... Media reader, 18 ... Recording medium, 21 ... Image database, 22 ... Search target database, 23 ... Error recognition probability database, 24 ... Program storage unit, 25 ... Input buffer, 26 ... Output buffer, 27 ... Search Number storage unit, 28 ... search candidate storage unit.

Claims (6)

A search object acquisition means for acquiring text-recognized text data as a search object;
For each character, a database for erroneous recognition in which characters that are likely to be erroneously recognized are stored as erroneously recognized characters,
An input means for entering a search term;
Character replacement means for replacing each character of the search term input by the input means with a misrecognized character stored in the erroneous recognition database;
A data search device comprising: search processing means for searching the text data by adding a search word after character replacement obtained by the character replacement means to the search word. In the database for misrecognition, a plurality of different misrecognized characters are stored together with a misrecognition rate in association with each character in advance.
2. The data search device according to claim 1, wherein the character replacement means replaces each character of the search word in order from a misrecognized character having a high error recognition rate.   The data according to claim 2, wherein the character replacement means replaces each character of the search word in order from the erroneous recognition character having a high recognition rate until a predetermined number of search results are obtained by the search processing means. Search device. Search candidate generation means for generating a predetermined number of search candidates by replacement with misrecognized characters by the character replacement means,
The data search apparatus according to claim 1 or 2, wherein the search processing means searches the text data by adding each search candidate generated by the search candidate generation means to the search word. Get text-recognized text data as search target,
For each character, a misrecognition database in which characters that are likely to be misrecognized are stored as misrecognized characters, and each character of the input search word is stored in the misrecognition database. Replace with letters,
A data search method, wherein the text data is searched by adding the search word after the character replacement to the search word. A data search program executed by a computer,
In the computer,
The ability to retrieve text data with character recognition as a search target;
For each character, a misrecognition database in which characters that are likely to be misrecognized are stored as misrecognized characters, and each character of the input search word is stored in the misrecognition database. The ability to replace characters,
A program for searching the text data by adding the search word after the character replacement to the search word.

Images