JP2011018108A - Device and program for correction of recognized character string - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a recognized character string correction device capable of correcting a character string recognized from an imaged document with high accuracy, and to provide a program for correction of the recognized character string.SOLUTION: The device includes: a correction information memory 31 for storing normal character string information for use in an object document and similar pattern information configured of the pertinent character information group of each similar pattern of characters and one representative character information selected from among the character information group; a character string comparison part 35 for comparing character string information acquired by converting character information in the character information group of the similar pattern in character information configuring normal character string information into the representative character information of the similar pattern with the character string information acquired by converting the acquired recognized character string information into the representative character information in the same way; and a character string correction part 36 for replacing the acquired recognized character string information with the normal character string information for correction, when the comparison results shows the identical character string information.

Description

  The present invention relates to a recognized character string correcting apparatus and a recognized character string correcting program for correcting a character string that is erroneously recognized when characters are recognized from an imaged document into an accurate character string.

  2. Description of the Related Art Conventionally, as a technique for recognizing characters from an imaged document by a computer, a technique using an apparatus such as an OCR or a text reader is known. Some of the functions of these devices are commercially available in software, and are widely used to convert newspaper articles, various specifications, books, etc. into text.

  By the way, in the character recognition process using these techniques, a character may be erroneously recognized, and it is necessary to correct it.

  As a technique for correcting a recognized character, there are techniques described in Patent Document 1 and Patent Document 2.

  Patent Document 1 uses a dictionary corresponding to a character code system created by contraction that represents a subset of similar characters that have a high possibility of misrecognition with fewer characters, and matches the target term. A technique for correcting errors at the single character level is described.

  In Patent Document 2, a word candidate is output together with a score indicating the certainty of the candidate word by recognizing the character in a term unit and collated with the term dictionary, and each term included in the term dictionary is Using the term discriminant table that calculates the discriminating degree of each term from the similarity of the terms and aggregates them, and corrects the term score according to the level of term discriminating degree for each term candidate, in order from the word candidate with the highest accuracy Techniques that can be shown are described.

JP-A-4-361393 Japanese Patent Laid-Open No. 4-68484

  By using the technique described in Patent Document 1 or Patent Document 2 described above, there is a high possibility that a misrecognized character string can be corrected to an accurate character string. However, when used in documents in a limited field, For example, when character strings used are used for character recognition such as limited forms and invoices, it has been desired to perform correction with higher accuracy. In particular, it has been desired to correct by accurately distinguishing “seo” and “n”, “tsu” and “shi”, etc., which are difficult to identify with a computer.

  Accordingly, an object of the present invention is to provide a recognized character string correction device and a recognized character string correction program capable of correcting a character string recognized from an imaged document with high accuracy.

  A recognition character string correction apparatus according to the present invention for solving the above problems is a recognition character string correction apparatus that corrects misrecognized characters in recognition character string information recognized by analyzing a document based on image image information. , Which is composed of preset character string information used for the document, corresponding character information group for each character similar pattern, and one representative character information selected from the character information group Correction information storage unit for storing similar pattern information, recognized character string information acquisition unit for acquiring the recognized character string information, and character information constituting the recognized character string information acquired by the recognized character string information acquisition unit Among them, a character string conversion unit that converts character information in a character information group of any similar pattern stored in the correction information storage unit into representative character information of the similar pattern; A character string obtained by converting character information in the character information group of any one of the similar patterns from character information constituting regular character string information stored in the correction information storage unit into representative character information of the similar pattern The character string comparison unit that compares the information with the character string information obtained by converting the recognized character string information by the recognized character string conversion unit, and as a result of comparison by the character string comparison unit, the regular character string information is When the converted character string information and the character string information obtained by converting the recognized character string information are the same, the recognized character string information acquired by the recognized character string information acquisition unit is stored in the correction information storage unit. And a character string correction unit that corrects the data by replacing with the stored regular character string information.

  In the recognized character string converting unit of the recognized character string correcting device, when the character information constituting the recognized character string information includes lowercase characters, the lowercase characters are converted into uppercase characters, and the character string comparing unit converts the correction information. Among the character information constituting the regular character string information stored in the storage unit, the character information in the character information group of any one of the similar patterns is converted into the representative character information of the similar pattern and the character information When the character string includes lowercase letters, the character string information obtained by converting the lowercase letters into uppercase letters may be compared with the character string information obtained by converting the recognized character string information by the recognized character string conversion unit.

  Further, the correction information storage unit of the recognized character string correction device converts character information included in the character information group of any one of the similar patterns from the character information constituting the regular character string information. The character string comparison unit further stores the normal conversion character string information converted into the representative character information, and the character string comparison unit stores the normal conversion character string information stored in the correction information storage unit and the recognized character string conversion unit. You may make it compare with the character string information into which column information was converted.

  The recognition character string correction program of the present invention is set in advance in a recognized character string correction device that corrects misrecognized characters in recognized character string information recognized by analyzing a document based on image image information. In addition, regular character string information used for the document, corresponding character information group for each character similar pattern, and similar pattern information composed of one representative character information selected from the character information group, The character information in the character information group of any one of the stored similar patterns among the character information constituting the acquired recognized character string information. Characters in the character information group of any one of the similar patterns among the function of converting to the representative character information of the similar pattern and the character information constituting the stored regular character string information A function for comparing the character string information obtained by converting the information into the representative character information of the similar pattern and the character string information obtained by converting the recognized character string information, and as a result of the comparison, the regular character string information is converted. When the obtained character string information is the same as the character string information obtained by converting the recognized character string information, a function of correcting the acquired recognized character string information by replacing it with the regular character string information; Is executed.

  According to the recognized character string correcting apparatus and the recognized character string correcting program of the present invention, the character string recognized from the imaged document can be corrected with high accuracy.

It is a block diagram which shows the structure of the character recognition system using the recognition character string correction apparatus by one Embodiment of this invention. It is a flowchart which shows operation | movement of the recognition character string correction apparatus by one Embodiment of this invention. It is explanatory drawing which shows an example of the regular character string information memorize | stored in the correction information storage part of the recognition character string correction apparatus by one Embodiment of this invention. It is explanatory drawing which shows an example of the similar pattern information memorize | stored in the correction information storage part of the recognition character string correction apparatus by one Embodiment of this invention. It is explanatory drawing explaining the process performed by the recognition character string conversion part of the recognition character string correction apparatus by one Embodiment of this invention, a regular character string conversion part, a character string comparison part, and a character string correction | amendment part.

  An embodiment of a character recognition system using a recognized character string correction apparatus of the present invention will be described with reference to the drawings.

<Configuration of Character Recognition System According to One Embodiment>
A character recognition system 1 according to the present embodiment recognizes characters from a printed document and generates a text file. As shown in FIG. 1, an image reading device 10, a character recognition device 20, a recognized character, and the like. A column correction device 30.

  The image reading apparatus 10 is a scanner or the like, and includes a reading control unit 11 that reads a document to be generated as a text file as image image data, and an image image data storage unit 12 that stores the read image image data.

  The character recognition device 20 is an OCR or the like. The character recognition control unit 21 recognizes character information by analyzing the image image data stored in the image image data storage unit 12 of the image reading device 10, and the recognized character information. And a recognized character string information storage unit 22 for storing recognized character string information, which is configured character string information.

  The recognized character string correction device 30 includes a correction information storage unit 31, a recognized character string information acquisition unit 32, a recognized character string conversion unit 33, a regular character string conversion unit 34, a character string comparison unit 35, and a character string correction. A unit 36, a text file generation unit 37, and a text file storage unit 38.

  The correction information storage unit 31 includes regular character string information used for the document set in advance, a corresponding character information group for each similar character pattern, and one representative character selected from the character information group Similar pattern information composed of information is stored.

  The recognized character string information acquisition unit 32 acquires the recognized character string information stored in the recognized character string information storage unit 22 of the character recognition device 20.

  In the character information group of any similar pattern in which each character information constituting the recognized character string information acquired by the recognized character string information acquisition unit 32 is stored in the correction information storage unit 31. The character information in the character information group of any similar pattern is converted into the representative character information of the similar pattern.

  The regular character string conversion unit 34 determines whether each character information constituting the regular character string information stored in the correction information storage unit 31 is in the character information group of any similar pattern, respectively. Character information in the character information group of any similar pattern is converted into representative character information of the similar pattern.

  In the character string comparison unit 35, the character string information whose recognition character string information is converted by the recognition character string conversion unit 33 and the character string information whose normal character string information is converted by the normal character string conversion unit 34 are the same. Compare whether or not.

  When the character string information converted from the recognized character string information and the character string information converted from the regular character string information are the same as a result of the comparison by the character string comparing unit 35, the character string correcting unit 36 recognizes Correction is performed by replacing the recognized character string information acquired by the character string information acquisition unit 32 with the corresponding regular character string information stored in the correction information storage unit 31.

  The text file generation unit 37 generates a text file composed of the character string information corrected by the character string correction unit 36.

  The text file storage unit 38 stores the text file generated by the text file generation unit 37.

<Operation of Character Recognition System According to One Embodiment>
Next, a process when a text file is generated and stored from the printed document X in the character recognition system 1 according to the present embodiment will be described.

  First, a document X, which is a printed document to be generated as a text file, is read as image image data by the reading control unit 11 of the image reading apparatus 10 and stored in the image image data storage unit 12.

  Next, the character recognition control unit 21 of the character recognition device 20 analyzes the image image data stored in the image image data storage unit 12 and recognizes character information. The character string information composed of the recognized character information is stored in the recognized character string information storage unit 22.

  Next, in the recognized character string correction device 30, a process of correcting a misrecognized character in the recognized character string information is performed. The correction process for the erroneously recognized character will be described with reference to the flowchart of FIG.

  In the correction information storage unit 31 of the recognized character string correction device 30, the regular character string information used for the document, the corresponding character information group for each similar character pattern, and the character information group are set in advance. And similar pattern information composed of one representative character information selected from the above.

  An example of regular character string information used for the document is shown in FIG. In the present embodiment, “software”, “solution 1”, and “middleware” are stored as regular character string information used for the document.

  FIG. 4 shows an example of similar pattern information composed of a corresponding character information group and one representative character information selected from the character information group for each similar pattern of characters. This similar pattern is generated by paying attention to similar characters that are difficult to identify with a computer. In the present embodiment, as similar pattern information, character information corresponding to each of the seven patterns of similar patterns A to G, respectively. A group and one representative character information selected from each character information group are stored.

  For example, in FIG. 4, “SO” and “N” which are similar character information are stored as the character information group of the similar pattern A, and “N” is selected as the representative character information in these character information groups. Stored. Also, similar character information “tsu” and “shi” are stored as the character information group of the similar pattern B, and “shi” is selected and stored as representative character information in these character information groups. Yes. Also, “—” and “—” that are similar character information are stored as character information groups of the similar pattern C, and “—” is selected and stored as representative character information in these character information groups. Yes. Also, similar character information “1”, “I”, “|”, and “!” Are stored as character information groups of the similar pattern D, and “1” is represented as representative character information in these character information groups. "Is selected and stored. Also, similar character information “;” and “:” are stored as character information groups of the similar pattern E, and “:” is selected and stored as representative character information in these character information groups. Yes. Further, similar character information “.”, “,” And “.” Are stored as character information groups of the similar pattern F, and “.” Is selected as representative character information in these character information groups. Stored. Further, similar character information “+” and “ten” are stored as the character information group of the similar pattern G, and “+” is selected and stored as the representative character information in these character information groups. Yes.

  When a process for correcting a misrecognized character is started in a state in which these regular character string information and similar pattern information are stored in the correction information storage unit 31, first, a character string information acquisition unit 32 performs a character search. The recognized character string information stored in the recognized character string information storage unit 22 of the recognition device 20 is acquired (S1). Here, it is assumed that “Narutsuso I” is acquired as the recognized character string information.

  Next, in the recognized character string conversion unit 33, each character information “n”, “li”, “u”, which constitutes the recognized character string information “Nunyusoso I” acquired by the recognized character string information acquisition unit 32. It is determined whether or not “-”, “tu”, “yo”, “so”, “I” are in the character information group of any similar pattern stored in the correction information storage unit 31, respectively. The character information in the character information group of any similar pattern is converted into the representative character information of the corresponding similar pattern as indicated by 51 in FIG. 5 (S2).

  Here, it is first determined that the character information “n” is in the character information group of the similar pattern A, and this character information “n” is converted into the representative character information “n” of the similar pattern A. Here, since the character information to be determined and the representative character information to be converted are the same, the character information does not actually change.

  Further, it is determined that the character information “-” is in the character information group of the similar pattern C, and this character information “-” is converted into the representative character information “-” of the similar pattern C. Again, since the character information to be determined and the representative character information to be converted are the same, the character information does not actually change.

  Further, it is determined that the character information “TSU” is in the character information group of the similar pattern B, and this character information “TSU” is converted into the representative character information “SH” of the similar pattern B.

  Further, it is determined that the character information “SO” is in the character information group of the similar pattern A, and this character information “SO” is converted into the representative character information “N” of the similar pattern A.

  Further, it is determined that the character information “I” is in the character information group of the similar pattern D, and this character information “I” is converted into the representative character information “1” of the similar pattern D.

  By converting each piece of character information in this way, the recognized character string information “Narutsuso I” is converted to “version 1”.

  Next, in the regular character string conversion unit 34, each character information constituting the regular character string information stored in the correction information storage unit 31 is in the character information group of any similar pattern, respectively. The character information that has been determined and was in the character information group of any similar pattern is converted into the representative character information of the corresponding similar pattern (S3).

  Here, first, among the regular character string information stored as shown in FIG. 3, each character information “So”, “F”, “G”, “W” constituting the first character string information “software” is displayed. It is determined whether or not “”, “e”, and “a” are in the character information group of any similar pattern stored in the correction information storage unit 31.

  As in the case of the recognized character string information described above, as a result of determination for each character information, the first regular character string information “software” is converted to “software”.

  Next, in the character string comparison unit 35, the character string information obtained by converting the recognized character string information in step S2 is the same as the character string information obtained by converting the first regular character string information in step S3. Whether or not is compared (S4).

  As a result of the comparison, the character string information “version 1” converted in step S2 is not the same as the character string information “software” converted in step S3 (“NO” in S5). The process proceeds to the next second regular character string information (S6).

  Returning to step S3, the character string conversion unit 34 uses the character information “So”, “Li”, “Yu”, “-”, “Sh” constituting the second normal character string information “Solution 1”. , “X”, “N”, and “1” are determined to be in the character information group of any similar pattern stored in the correction information storage unit 31.

  As a result of the determination for each character information in the same manner as the recognized character string information described above, the second regular character string information “solution 1” is converted to “solution 1” as indicated by 52 in FIG. .

  Next, in the character string comparison unit 35, the character string information obtained by converting the recognized character string information in step S2 is the same as the character string information obtained by converting the second regular character string information in step S3. Whether or not is compared (S4).

  As a result of the comparison, it is determined here that the character string information “version 1” converted in step S2 is the same as the character string information “solution 1” converted in step S3 (S5). "YES")

  If it is determined that the character string information converted in step S2 is the same as the character string information converted in step S3, the character string correction unit 36 obtains recognized character string information as shown at 53 in FIG. Correction is performed by replacing the recognized character string information “Narutsuso I” acquired by the unit 32 with the corresponding second regular character string information “Solution 1” (S7).

  Then, a text file composed of the character string information corrected by the character string correction unit 36 is generated by the text file generation unit 37 (S8) and stored in the text file storage unit 38 (S9).

  According to the above embodiment, in a limited field document in which character information to be used is determined to some extent, a character string recognized from a printed document is corrected with high accuracy to generate an accurate text file. It becomes possible to do.

  Although the above has been described based on the best mode of the present invention, it should not be understood that the description and the drawings, which form a part of this disclosure, limit the present invention. From this disclosure, various alternative embodiments, examples, and operational techniques will be apparent to those skilled in the art.

  For example, in the present embodiment, a case has been described in which each regular character string information is sequentially converted and compared each time correction processing is performed in the character string comparison unit 35, but each regular character string is compared. Character string information obtained by converting information based on similar pattern information may be stored in advance in the correction information storage unit 31 and used for comparison processing.

  In addition, when the recognition character string conversion unit 33 and the regular character string conversion unit 34 perform conversion processing based on the stored similar pattern information as shown in FIG. 4, uppercase letters such as “tsu” and “tsu” are used. By converting the conversion of lowercase letters to uppercase letters, it is possible to reduce the amount of information stored in similar pattern information and correct it correctly even if the size is misrecognized. And the correction accuracy can be improved.

  It goes without saying that the present invention includes various embodiments not described herein. Accordingly, the technical scope of the present invention is defined only by the invention specifying matters according to the scope of claims reasonable from the above description.

DESCRIPTION OF SYMBOLS 1 ... Character recognition system 10 ... Image reading apparatus 11 ... Reading control part 12 ... Image image data memory | storage part 20 ... Character recognition apparatus 21 ... Character recognition control part 22 ... Recognition character string information storage part 30 ... Recognition character string correction | amendment apparatus 31 ... Correction information storage unit 32 ... Recognition character string information acquisition unit 33 ... Recognition character string conversion unit 34 ... Regular character string conversion unit 35 ... Character string comparison unit 36 ... Character string correction unit 37 ... Text file generation unit 38 ... Text file storage unit

Claims (4)

In a recognized character string correction device that corrects misrecognized characters in recognized character string information recognized by analyzing a document based on image image information,
Pre-set regular character string information used for the document, corresponding character information group for each character similar pattern, and similarity composed of one representative character information selected from this character information group A correction information storage unit for storing pattern information;
A recognized character string information acquisition unit for acquiring the recognized character string information;
Among the character information constituting the recognized character string information acquired by the recognized character string information acquiring unit, character information in the character information group of any similar pattern stored in the correction information storage unit A recognized character string conversion unit for converting into representative character information of a pattern;
Character information that is included in the character information group of any one of the similar patterns out of character information that constitutes regular character string information stored in the correction information storage unit, and is converted into representative character information of the similar pattern A character string comparison unit that compares column information with character string information obtained by converting the recognized character string information by the recognized character string conversion unit;
When the character string information converted from the regular character string information and the character string information converted from the recognized character string information are the same as a result of the comparison by the character string comparison unit, the recognized character string information A character string correction unit that corrects the recognized character string information acquired by the acquisition unit by replacing it with the regular character string information stored in the correction information storage unit;
A recognition character string correction apparatus comprising: Further, in the recognized character string conversion unit, when lowercase letters are included in the character information constituting the recognized character string information, the lowercase letters are converted into uppercase letters,
In the character string comparison unit, the character information included in the character information group of any one of the similar patterns among the character information constituting the regular character string information stored in the correction information storage unit. When the character information includes lower case characters, the character string information obtained by converting the lower case characters to upper case is compared with the character string information obtained by converting the recognized character string information by the recognized character string conversion unit. The recognition character string correction device according to claim 1, wherein The correction information storage unit converts the character information included in the character information group of any one of the similar patterns from the character information constituting the regular character string information into the representative character information of the similar pattern. Store more string information,
The character string comparison unit compares the normal conversion character string information stored in the correction information storage unit with the character string information obtained by converting the recognition character string information in the recognition character string conversion unit. The recognition character string correction device according to claim 1 or 2. To a recognized character string correction device that corrects misrecognized characters in recognized character string information recognized by analyzing a document based on image image information,
Pre-set regular character string information used for the document, corresponding character information group for each character similar pattern, and similarity composed of one representative character information selected from this character information group A function for storing pattern information;
A function of acquiring the recognized character string information;
A function of converting character information in the character information group of any one of the similar patterns out of the character information constituting the acquired recognized character string information into representative character information of the similar pattern;
Character information in the character information group of any one of the similar patterns out of character information constituting the stored regular character string information, character string information converted into representative character information of the similar pattern, and the recognition A function for comparing the character string information with the converted character string information;
As a result of the comparison, when the character string information converted from the regular character string information is the same as the character string information converted from the recognized character string information, the acquired recognized character string information is A function to correct by replacing with regular character string information,
Recognition character string correction program that executes

Images