Information processing device, information processing method, program, and computer-readable storage medium

Abstract

To reduce erroneous conversion of character strings that have been erroneously recognized in character recognition processing on a document.SOLUTION: An information processing device 1 includes a storage unit 20 and a processing unit 10. The storage unit 20 stores a list 200 of conversion rules 201 in which each of a plurality of words having different numbers of characters is associated with a conversion target character string for identifying a character string including an erroneously recognized character as a conversion target. The processing unit 10 compares the conversion rules 201 in ascending order of the number of characters in the list 200 with character strings obtained by character recognition processing on a document, and sequentially converts the character strings that match the conversion target character strings in the character strings obtained by the character recognition processing into the corresponding words.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device, an information processing method, a program, and a computer-readable storage medium, and relates to a technique for converting characters that are erroneously recognized in character recognition processing of image data of a document, for example. [Background technology]

[0002] A method is known in which optical character recognition (OCR) processing is applied to image data obtained by scanning a document using a device such as a camera or scanner, and the recognized character data (e.g., text data) is output. Such character recognition processing can be used, for example, to digitize documents printed on paper, promoting paperless operations. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 9-138836 Summary of the Invention [Problem to be solved by the invention]

[0004] In OCR processing, misrecognition of characters can occur due to factors such as the printing state of characters in image data (hereinafter referred to as "document image") obtained by reading a document using a device such as a camera or scanner, or the overlap of characters with a background image, resulting in the generation of incorrect character data.

[0005] As an example of a method for correcting misrecognized character strings, a conversion list that defines the correspondence between correct character strings and misrecognized character strings is known (for example, Patent Document 1). In this method, the conversion list is used to mechanically convert misrecognized character strings into correct character strings in post-OCR processing.

[0006] However, for example, when the conversion list as described above is mechanically applied to each of words with different numbers of characters, erroneous conversion may occur. For example, when a first word (character string) in which misrecognition occurred is converted into a correct character string, a word with a larger number of characters that includes the correct character string may be converted erroneously.

[0007] For example, if a conversion list that associates the correct word (string) "vector" with the misrecognized string "pector" is applied to the misrecognized string "pector," it will be converted to the correct string "vector."

[0008] However, if the input string contains the correct word "spectrum," the same conversion list will be applied automatically, resulting in the "spectrum" being incorrectly converted to "spectrum vector" because the "spectrum" part will be converted to "vector." This may ultimately require, for example, visual confirmation and correction of the conversion results.

[0009] One of the objectives of the present disclosure is to reduce erroneous conversion of character strings that are misrecognized in character recognition processing of a document. [Means for solving the problem]

[0010] An information processing device according to one aspect of the present disclosure includes a memory unit that stores a list of conversion rules that associates a plurality of words having different numbers of characters with a conversion target string for identifying a string containing a misrecognized character as a conversion target, and a processing unit that compares the conversion rules in the list in ascending order of the number of characters against a string obtained by character recognition processing on a document, and sequentially converts a string in the string obtained by the character recognition processing that matches the conversion target string into a corresponding word. [Effects of the Invention]

[0011] According to the present disclosure, character strings obtained by character recognition processing on a document are converted into corresponding words by checking conversion rules in order, starting with words with the smallest number of characters, thereby reducing the erroneous conversion of character strings that have been misrecognized. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 10 is a diagram illustrating an example of a character string conversion process according to an embodiment. [Figure 2] 1 is a block diagram illustrating an example of the configuration of an information processing device according to an embodiment. [Figure 3] 1A to 1C are diagrams showing examples of conversion rules according to an embodiment. [Figure 4] FIG. 10 is a diagram illustrating an example of log data of a conversion process according to an embodiment. [Figure 5] FIG. 10 is a diagram illustrating an example of generating a conversion target character string according to an embodiment. [Figure 6] FIG. 10 is a diagram showing an example of the format of one record among a plurality of records constituting a list of conversion rules according to an embodiment. [Figure 7] FIG. 2 illustrates an example of a data structure that defines a list of conversion rules according to an embodiment. [Figure 8] 10 is an exemplary flowchart illustrating a character string conversion process according to an embodiment. [Figure 9] FIG. 10 is a diagram illustrating an example of a character string conversion process using a conversion rule according to an embodiment. [Figure 10] 10 is an exemplary flowchart illustrating a process for generating a conversion rule according to an embodiment. [Figure 11] 7A is a diagram showing an example of a conversion rule record generated in one embodiment, and FIG. 7B is a diagram showing an example in which the record shown in FIG. 7A is added to the list of conversion rules shown in FIG. [Figure 12] FIG. 11(b) is a diagram showing an example of a character string conversion process using the list of conversion rules shown in FIG. [Figure 13]10(a) and 10(b) are diagrams showing an example of a list of conversion rules according to a modified example. [Figure 14] 10 is an exemplary flowchart illustrating a process for generating a conversion rule according to a modified example. [Figure 15] 10 is an exemplary flowchart illustrating a character string conversion process according to a modified example. [Figure 16] 3 is a diagram illustrating an example of a hardware configuration of the information processing device illustrated in FIG. 2. FIG. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, embodiments and examples will be described in detail with reference to the drawings. The same or similar reference numerals indicate the same or similar elements, and repeated explanations may be omitted. When numerical values are described in the following description, the numerical values are merely examples, and other numerical values may be used in addition or instead.

[0014] <Summary> 1 is a diagram illustrating an example of a character string conversion process according to an embodiment. In FIG. 1, the character string "spectrum" included in an input character string is converted to "vector" by applying conversion rule A, and the character string "spectrum" that includes the converted character string "vector" as a part of it is converted to "spectrum" by applying conversion rule B.

[0015] Conversion rule A associates the string "[bepepe]kutor", which contains a square bracket [], a regular expression, at the first character position, as the target string for conversion for the four-character correct string "vector". Conversion rule B associates the string "s[bebepe]kutor", which contains a square bracket [], a regular expression, at the second character position, as the target string for conversion for the five-character correct string "spectrum".

[0016] <regular expression> A regular expression is a method for expressing multiple patterns of strings in a single format. Square brackets [] (also known as metacharacters), which are one of the regular expressions, are used to express multiple patterns of strings containing any one of the multiple characters enclosed in the square brackets [] in a single string format.

[0017] For example, a string expressed as [ABC]DE represents three possible string patterns: "ADE", "BDE", and "CDE", where the character in the first character position is either "A", "B", or "C".

[0018] Here, assume that the katakana "be" at the first character position in the correct answer character string "vector" can be misrecognized in three ways: as the hiragana "be," as the katakana "pe," and as the hiragana "pe." In this case, three patterns of character strings (in other words, misrecognized character strings) that contain one of the three misrecognition candidate characters [bepepe] at the first character position are represented by a single character string "[bepepe]kutor."

[0019] Similarly, assume that the katakana character "pe" at the second character position of the correct character "spectrum" can be misrecognized in three ways: as the katakana character "be," as the hiragana character "be," and as the hiragana character "pe." In this case, three patterns of character strings containing one of the three misrecognition candidate characters [bebepe] at the second character position are represented by a single character string "S[bebepe]kutor."

[0020] Therefore, conversion rule A means that in the input string, a string that matches one of the three misrecognition candidate characters [bepepe] at the first character position of the four-character "vector" is converted into the correct string "vector."

[0021] Conversion rule B is applied to strings with a larger number of characters than conversion rule A, and means that any string that matches one of the three misrecognized candidate characters [bebepe] at the second character position of the five-character "spectrum" will be converted to the correct string "spectrum."

[0022] Thus, if the input string contains the word "spectrum," application of conversion rule A will convert the string "spectrum" to "vector," but subsequent application of conversion rule B will convert it to the correct string word "spectrum."

[0023] In this way, in the conversion process illustrated in Figure 1, conversion rules A and B are mechanically and sequentially applied in ascending order of the number of characters in the correct string, without confirming or determining whether the string (word) is a correct conversion or an incorrect conversion.

[0024] In other words, the conversion process allows for cases where a misrecognized character string is converted to a correct character string, as well as cases where a correct character string is converted to an incorrect character string. Even if the latter conversion occurs, there is a chance that it will eventually be converted (or corrected) into a word with a correct character string by later applying another conversion rule for a word with a longer number of characters that includes the misconverted character string.

[0025] "There is room" means that, since it is unrealistic to prepare comprehensive conversion rules in advance for all words with all numbers of characters, it is acceptable for there to be cases where incorrect conversions cannot be corrected even by sequentially applying the conversion rules in ascending order according to the number of characters described above.

[0026] In the embodiment described below, an example will also be described in which new (or additional) conversion rules are automatically generated for strings in which erroneous conversions remain, and the list of conversion rules is updated, so that the probability of such cases occurring can be adaptively reduced.

[0027] <Explanation of terms etc.> To facilitate understanding of the embodiments described below, some terms used in this specification will be explained below.

[0028] A "document" is a document containing multiple characters or multiple strings of characters (e.g., words), and a "document image," as already mentioned, means image data obtained by reading a document with a device such as a camera or scanner.

[0029] "Characters" include any characters that can be used in a document, including, but not limited to, kanji, hiragana, katakana, alphabets, symbols, and numbers. In this embodiment, a document written in Japanese will be used as an example. In character recognition processing, misrecognition of characters between these characters can occur.

[0030] Converting (or substituting) a "character" or a "character string" with another character or character string may be understood as converting (or substituting) the character code corresponding to the character or the character code set corresponding to the character string.

[0031] A "correct character string" or a "correct word" refers to a correct character string or word that is not misrecognized, and a "correct character" refers to the individual characters that make up a correct character string. Note that a correct character string may be understood to be, for example, a predetermined character string as a word. A "misrecognized character candidate" is a concept that includes both an "misrecognized character" that has been confirmed to have been misrecognized in the character recognition processing results, and a candidate character that is expected to be misrecognized in the character recognition processing.

[0032] A "conversion rule" is data that associates (or relates) a correct string (or correct word) with a "string to be converted," and a string in the input string that matches (or matches) the "string to be converted" is converted into the corresponding "correct string."

[0033] The "conversion target character string" refers to a character string for identifying a character string containing a misrecognized character in the character recognition processing result as a conversion target. For example, the "conversion target character string" is matched or searched in the character recognition processing result, and a character string that matches the "conversion target character string" is identified as a character string containing a misrecognized character. The character string identified in this way is the target for conversion to a correct character string. Note that the term "identification" may be interchangeably interpreted as other terms such as "identification" or "detection" depending on the context, for example.

[0034] If there is one candidate (or pattern) for the "conversion target string," for example, if there is one pattern of a string that may be misrecognized as the correct string (misrecognized string), the "conversion target string" will match the one pattern of the "misrecognized string."

[0035] On the other hand, if there are multiple candidates for the "conversion target string," the "conversion target string" can be expressed as a single string format using one or more regular expressions. By using regular expressions, it is not necessary to prepare individual conversion rules for multiple candidates. Therefore, the number of conversion rules that need to be prepared for the conversion process can be reduced.

[0036] In the embodiment described below, the "plurality of candidates" (in other words, the plurality of character string patterns) may or may not include the pattern of the correct character string in addition to the pattern of the incorrectly recognized character string.

[0037] A case where the "multiple candidates" do not include the pattern of the correct string is, for example, a case where "one character" in the correct string can be misrecognized as multiple different characters. In this case, the "string to be converted" represents multiple patterns of misrecognized strings using one string using one regular expression corresponding to "one character." The "string to be converted" in conversion rules A and B shown in Figure 1 corresponds to this case.

[0038] On the other hand, when the "multiple candidates" include the pattern of the correct answer string, each of the "multiple characters" in the correct answer string may be misrecognized as one or more different characters. In this case, the "string to be converted" represents a string pattern that includes both the pattern of the misrecognized string and the pattern of the correct answer string, using a single string that uses multiple regular expressions corresponding to each of the "multiple characters." Note that the conversion rule in this case means that it is permissible to convert a correct answer string into the same correct answer string. The technical significance of this will be explained later.

[0039] The meanings of terms etc. have been explained above. Below, an embodiment that realizes the character string conversion process as shown in FIG. 1 will be explained.

[0040] <Configuration example> 2 is a block diagram showing an example of the configuration of an information processing device 1 according to an embodiment. The information processing device 1 may be understood to function as a character data processing device that processes character data that is a result of character recognition processing by OCR processing.

[0041] 2, the information processing device 1 includes, for example, a processing unit 10, a storage unit 20, and a display unit 30. The processing unit 10 performs processing on an input character string, which is an example of character data (for example, conversion processing of the character string on a word-by-word basis).

[0042] The storage unit 20 stores, for example, information that is written or read as appropriate by the processing unit 10 in accordance with the processing or operation of the processing unit 10. The storage unit 20 may be provided inside the information processing device 1 as illustrated in Fig. 2, or may be provided outside the information processing device 1 via a predetermined interface.

[0043] For example, the storage unit 20 may be provided in an external device that can be accessed by the processing unit 10 via a network interface. The network interface may be either a wireless or wired interface.

[0044] The information stored in the memory unit 20 may include, for example, a program or data. The data may include, for example, a list 200 of conversion rules 201-1 to 201-n (n is an integer of 2 or more) prepared separately for words with different numbers of characters (hereinafter sometimes abbreviated as "conversion list 200"). For an example of the data structure of the conversion list 200, refer to FIG. 7 which will be described later.

[0045] When not distinguishing each of the conversion rules 201-1 to 201-n, it is abbreviated as "conversion rule 201". Further, the data stored in the memory unit 20 may include log data 203. Regarding the log data 203, refer to FIG. 4 which will be described later.

[0046] The display unit 30 displays information indicating the process or result of the process by the processing unit 10, or information read from the memory unit 20 by the processing unit 10, in accordance with the display control of the processing unit 10. Exemplarily, the conversion list 200 (or conversion rule 201), or the log data 203 may be displayed on the display unit 30.

[0047] <Regarding the conversion rule> FIGS. 3(a) to 3(c) are diagrams showing examples of the conversion rules 201-1 to 201-3 according to the present embodiment, respectively.

[0048] <Case where the conversion target string is represented using a plurality of regular expressions> FIG. 3(a) shows an example of the conversion rule 201 for a case where there is one or more misrecognition candidate characters for each of a plurality of characters (for example, the first and third characters) in the correct character string of the word "vector".

[0049] FIG. 3(a) shows a case where, in the correct character string "vector", the first katakana "ベ" can be misrecognized as the hiragana "べ", the katakana "ペ", and the hiragana "ぺ", and the third katakana "ト" can be misrecognized as the kanji "卜".

[0050] In this case, the string to be converted is represented by a string in which each of a plurality of characters that may be misrecognized in the correct string is replaced with a square bracket [] of a regular expression, and each of the square brackets [] includes not only the misrecognition candidate characters but also the correct characters. In the following, the square bracket [] of the regular expression may be conveniently referred to as "regular expression []".

[0051] For example, as shown in FIG. 3(a), the string to be converted that is converted to the correct string "vector" by the application of the conversion rule 201-1 is represented by a string including the regular expression [] at the first character position and the regular expression [] at the third character position.

[0052] The regular expression [] at the first character position includes a total of four characters, namely, the hiragana character "べ", the katakana character "ペ", and the hiragana character "ぺ" which are three misrecognition candidate characters, and the katakana character "ベ" which is the correct character. The regular expression [] at the third character position includes a total of two characters, namely, the kanji character "卜" which is one misrecognition candidate character and the katakana character "ト" which is the correct character.

[0053] Therefore, in the conversion rule 201-1, there are 4×2 = 8 patterns of candidates in the string to be converted, and among these 8 patterns, in addition to the pattern of the misrecognized string, there is also a pattern corresponding to the correct string.

[0054] In other words, when the string to be converted is a string represented by the regular expression [] at a plurality of character positions, the rule is to include not only the pattern for converting the misrecognized string to the correct string but also the pattern for converting the correct string to the correct string in the regular expression []. This is because if the correct characters are not added to the regular expression [], a case where the conversion to the correct string is not supported may occur.

[0055] For example, in the conversion rule 201-1, when the string to be converted is represented as "[べぺペ]ク[ト卜]ル" which does not include the katakana character "ベ" which is the correct character in the regular expression [] for the first character position, "ベク卜ル" including the kanji character "卜" will not be the conversion target.

[0056] Therefore, if there are multiple characters (positions) in a correct character string that may be misrecognized, the regular expression [ ] for each character position will include the misrecognition candidate character and the correct character.

[0057] <The case where the string to be converted is expressed using one regular expression> FIG. 3(b) shows an example of the conversion rule 201-2 for a case where there are multiple erroneously recognized candidate characters for one character in the correct character string.

[0058] For example, the katakana "be" at the beginning of "Bayes' method" can be misrecognized in three ways: as the hiragana "be," as the katakana "pe," and as the hiragana "pe." Note that conversion rules A and B shown in Figure 1 also correspond to this case.

[0059] In this case, the conversion target string can be represented by a string in which one character that may be misrecognized in the correct string is replaced with the regular expression []. For example, in conversion rule 201-2, the conversion target string is represented as the string "[bepepe]ishou" in which the first character of the correct string is replaced with one regular expression [] that includes three possible misrecognition candidates.

[0060] Unlike the example shown in Figure 3(a), the correct character does not have to be included in this regular expression []. For example, the correct character, katakana "be", does not have to be included in the regular expression [] at the beginning of "Bayes' method".

[0061] If the correct character is included in the regular expression [] for the first character, such as [bebepepe], then the correct string "Bayes' method" may be converted into the same correct string, which may result in unnecessary conversion. Therefore, if the string to be converted can be expressed using a single regular expression [], not including the correct character in that regular expression [] will prevent unnecessary conversion and improve the efficiency of the conversion process.

[0062] <Cases where regular expressions are not used in the conversion target string> Next, FIG. 3(c) shows an example of the conversion rule 201-3 for the case where there is one candidate for a misrecognized character for one character in the correct character string.

[0063] In this case, since the correct character string and the misrecognized character string have a one-to-one relationship, the conversion target string can be expressed without using the regular expression []. For example, for the three-character correct character string "projection method", when there is one candidate for misrecognition of the character "射" at the second character position being "彭", the misrecognized character string is one pattern of "射彭法". Therefore, in the conversion rule 201-3, the conversion target string is the string "射彭法" without using the regular expression [].

[0064] <Configuration example of the processing unit 10> Referring back to FIG. 2, the processing unit 10 illustratively includes a conversion processing unit 102 that sequentially executes string conversion processing by applying (or collating) the conversion rule 201 to the input string in ascending order of the number of characters, and a conversion rule generation processing unit 104 that generates the conversion rule 201.

[0065] <Configuration example of the conversion processing unit 102> The conversion processing unit 102 illustratively includes an acquisition unit 122, a collation unit 124, a conversion unit 126, and a log output unit 128.

[0066] The acquisition unit 122 acquires, for example, a character string that is the result of character recognition (e.g., OCR) of a document image. The document image is generated, for example, by reading a document such as an academic paper, a corporate technical report, or a computer software manual with an optical reading device such as a scanner.

[0067] The character recognition processing may be performed in the processing unit 10 through reading of a document by a reading device connected to the information processing device 1, or may be performed in a device (e.g., a server, etc.) located outside the information processing device 1.

[0068] Furthermore, the character string resulting from the character recognition process may be obtained, for example, from a document file (e.g., a text file) obtained by the character recognition process and stored in the storage unit 20. The document file may be provided to the information processing device 1 via a computer-readable storage medium or a network. The network may be a wired or wireless network, or a mixed network of wired and wireless networks.

[0069] For example, the matching unit 124 applies the conversion rules 201 by number of characters stored in the memory unit 20 to the input string acquired by the acquisition unit 122 on a word-by-word basis in ascending order of the number of characters, and checks whether there is a string that matches the string to be converted in the conversion rules 201.

[0070] For example, when a character string that matches the conversion target character string is found in the input character string as a result of matching by the matching unit 124, the conversion unit 126 converts the character string into the correct character string associated with the conversion rule 201. The conversion unit 126 and the matching unit 124 may be functionally integrated into one block.

[0071] The log output unit 128 generates, for example, log data 203 of the conversion process by the conversion unit 126 and outputs it to the storage unit 20. As a non-limiting example, the log output unit 128 can output the log data 203 in the form of a format 400 as shown in FIG. 4 by outputting the result of the conversion process to which the conversion rule 201 is applied for each original document image of the input character string and number (No.) information that manages the conversion rule 201.

[0072] 4 includes document number information 402 for identifying the document that is the source of the document image, record number (No.) information 404 of the conversion rule 201, an erroneously recognized character string 406, a converted character string 408, surrounding character information before conversion 410, and surrounding character information after conversion 412. Note that the record number information 404 for managing (or identifying) the conversion rule 201 may be understood to correspond to the record number in the conversion list 200, which will be described later.

[0073] Record number information 404 in log data 203 is an example of information indicating which conversion rule 201 in conversion list 200 was applied to the match. Misrecognized character string 406 corresponds to, for example, a character string in an input character string that matches a character string to be converted in conversion rule 201. Converted character string 408 corresponds to, for example, a character string converted into a correct character string by applying the conversion rule 201.

[0074] The surrounding character information 410 and 412 are information indicating, for example, the characters included within a range of several to several tens of characters before and after the character string identified as the conversion target in the input character string, respectively, before and after conversion. Fig. 4 shows an example in which the character string identified as the conversion target is highlighted by bolding and underlining to make it easier to see. The highlighting may alternatively or additionally be performed in other ways, such as by color coding the character string or by color coding the highlights added to the character string.

[0075] The format 400 illustrated in FIG. 4 may be understood to correspond to an example of a display screen when the log data 203 is displayed on the display unit 30, or may be understood to correspond to an example of a printout when the log data 203 is printed.

[0076] Such log data 203 visualizes the process and contents of the conversion process, making it possible to clarify which conversion rule 201 of which record number in the conversion list 200 was used to execute the conversion process. Therefore, the validity of the conversion process contents can be easily confirmed, and appropriate operation and management of the conversion list 200 can be achieved.

[0077] <Configuration Example of Conversion Rule Generation Processor 104> 2, conversion rule generation processing unit 104 includes, for example, character string generation unit 142, conversion rule generation unit 144, and management unit 146.

[0078] The character string generation unit 142, for example, uses the character recognition correction data 502 to identify characters that can be misrecognized for each correct character of a word, and generates the above-described conversion target character string based on the identified characters that can be misrecognized. The characters that can be misrecognized can be identified, for example, by checking samples of character data for which the conversion process has ended or the log data 203 of the conversion process.

[0079] As a non-limiting example, as shown in FIGS. 2 and 5, the character string generation unit 142 generates a conversion target character string 506 including candidates for misrecognized character strings based on the character recognition correction data 502 and the correct character string (correct word) 504 of the word in which character misrecognition has occurred.

[0080] The character recognition correction data 502 is, for example, data that associates, on a character-by-character basis, the correct characters constituting a word with characters (misrecognized characters) that can be misrecognized for the correct characters in the character recognition process. FIG. 5 shows, as a non-limiting example, character recognition correction data 502 in a table format, and examples in which each of the capital letter "I", the katakana "ト", the katakana "ベ", the katakana "ペ", and the Chinese character "影" has been misrecognized (or can be misrecognized) are shown.

[0081] In this case, in the character recognition correction data 502, the number "1" is associated as a misrecognized character with the capital letter "I" which is the correct character, and the Chinese character "卜" is associated as a misrecognized character with the katakana "ト" which is the correct character.

[0082] Similarly, in the character recognition correction data 502, the hiragana "べ", the katakana "ペ", and the hiragana "ぺ" are associated as misrecognized characters with the katakana "ベ" which is the correct character. Also, in the character recognition correction data 502, the katakana "ベ", the hiragana "べ", and the hiragana "ぺ" are associated as misrecognized characters with the katakana "ペ" which is the correct character. Further, in the character recognition correction data 502, the Chinese character "彭" which is another Chinese character is associated as a misrecognized character with the Chinese character "影" which is the correct character.

[0083] Regarding other characters not illustrated in FIG. 5, similarly, correct characters and misrecognized characters are registered in the character recognition correction data 502.

[0084] The character recognition correction data 502 and the correct words 504 may be prepared in advance, for example, based on samples of character data for which the above-described conversion processing has ended or log data 203 of the conversion processing. These character recognition correction data 502 and correct words 504 may be input to the character string generation unit 142 separately from the input character string to the acquisition unit 122, or may be stored in the storage unit 20.

[0085] Based on the character recognition correction data 502 and the correct character string 504 of the word (for example, "index vector"), the character string generation unit 142 generates, for example, a character string 506 (for example, "index [bebepepe] vect[tobo]tor") for specifying the object to be converted to the correct word 504.

[0086] This character string 506 corresponds to the case illustrated in FIG. 3(a). As illustrated in the lower part of FIG. 5, in the regular expressions [] respectively assigned to the 7th and 10th character positions from the beginning, in addition to the misrecognized characters, the correct characters are included.

[0087] For example, in the regular expression [] at the 7th character position, the katakana "be" which is the correct character, the hiragana "be", the katakana "pe", and the hiragana "pe" which are misrecognized characters are included. Also, in the regular expression [] at the 10th character position, the katakana "to" which is the correct character and the Chinese character "bo" which is the misrecognized character are included.

[0088] Regarding other words not illustrated in FIG. 5, similarly, based on the character recognition correction data 502 and the correct words 504, a conversion target character string 506 for the correct word 504 is generated.

[0089] Returning to FIG. 2, the conversion rule generation unit 144 generates, for example, a conversion rule 201 that associates the conversion target character string 506 generated by the character string generation unit 142 with the correct word 504.

[0090] The management unit 146 manages the conversion rules 201 generated by the conversion rule generation unit 144, for example, by adding them to the conversion list 200 in the storage unit 20. For example, the management unit 146 may manage the conversion rules 201 by sorting them in ascending order of the number of characters in the correct words.

[0091] Therefore, for example, the management unit 146 manages the conversion list 200 in units of records in which a record number based on the number of characters of the correct word and information indicating the number of characters of the correct word are assigned to the conversion rule 201.

[0092] In other words, one record (or one entry) constituting the conversion list 200 may be defined, for example, by a format that associates a number based on the number of characters in the correct word, the correct word, the string to be converted, and the number of characters in the correct word.

[0093] Fig. 6 shows an example of a format 600 of one of the multiple records that make up the conversion list 200. As a non-limiting example, Fig. 6 shows a record of the conversion rule 201 in which the correct answer character string "vector" is associated with the conversion target character string "[bebepepe]ku[totto]ru" expressed using the regular expression [].

[0094] 6, the number of the conversion rule 201, i.e., the record number (No.), is, as a non-limiting example, a seven-digit number. The first two digits represent the number of characters in the correct answer string (e.g., 04), and the following five digits represent a number (which may be called an "incremental value") that is incremented by one (+1) each time a record (or entry) is added. Note that the record number is not limited to seven digits, and may be any number of digits as long as it is in the form of a number based on the number of characters in the correct answer string.

[0095] The management unit 146 uses such record numbers as a key to sort the records in ascending order of record numbers, thereby updating the conversion list 200 so that records of conversion rules 201 with fewer characters in the correct string are positioned higher in the conversion list 200. Therefore, the management unit 146 can easily manage the conversion list 200 by, for example, sorting the conversion rules 201 in the conversion list 200 in ascending order of the number of characters in the correct word.

[0096] Fig. 7 shows an example of the data structure of the conversion list 200 in which the records of the conversion rules 201 are sorted in ascending order of the number of characters in the correct answer character string. Fig. 7 shows an example in which the records of the conversion rules 201 for each of the three correct answer character strings, "vector," "spectrum," and "bias vector," are sorted in ascending order of record numbers whose first two digits include a number indicating the "number of characters" (e.g., 04, 05, 08). In Fig. 7, the first two digits of the record numbers corresponding to the "number of characters" are shown in bold for convenience (the same applies hereinafter).

[0097] Since the conversion list 200 is managed in ascending order of the number of characters in the correct words, for example, the conversion processing unit 102 can apply the conversion rules 201 in ascending order of the number of characters in the correct words by reading the records in the conversion list 200 in ascending order of the record numbers. Therefore, for example, address control for reading the conversion rules 201 from the storage unit 20 can be simplified, and the conversion process using the conversion rules 201 can be speeded up or made more efficient. Note that it is not essential that the conversion list 200 be sorted in ascending order of the number of characters in the correct strings. For example, the processing unit 10 can control the order in which the conversion rules 201 are read from the conversion list 200, thereby realizing the application of the conversion rules 201 in ascending order.

[0098] When performing conversion processing using the conversion list 200 illustrated in FIG. 7, for example, the misrecognized character string "bias spectrum" in the input character string is converted into the correct character string "bias vector" by performing conversion processing in the following order.

[0099] (1) By applying the conversion rule 201 of the record number = "0490068", the character string "vector" that matches the conversion target character string "[bebepepe]ku[totr]ru" is converted to "vector".

[0100] (2) Next, by applying the conversion rule 201 of the record number = "0590069", the character string "specter" that matches the conversion target character string "S[bebepepe]ku[totr]ru" is converted to "spectrum".

[0101] (3) Furthermore, by applying the conversion rule 201 of the record number = "0890070", the character string "bias spectrum" that matches the conversion target character string "bias [bebepepe] ku [tottor] ru" is converted into "bias vector".

[0102] Note that the above-described character string generating unit 142, conversion rule generating unit 144, and management unit 146 may be functionally integrated into one block in part or in whole.

[0103] <Processing example> An example of processing by the information processing device 1 having the above-described configuration will be described below, with the character string conversion processing by the conversion processing unit 102 and the generation processing of the conversion rule 201 by the conversion rule generation processing unit 104 being separately described.

[0104] <Example of string conversion processing> 8 is a flowchart showing an example of processing performed by the information processing device 1, focusing on character string conversion processing. As shown in FIG. 8, the processing unit 10 (e.g., the conversion processing unit 102) of the information processing device 1 acquires an input character string in the acquisition unit 122 (S802), and passes the acquired character string to the collation unit 124.

[0105] Furthermore, the conversion processing unit 102 reads, for example, the first record of the conversion list 200 from the storage unit 20, and passes the conversion rule 201 of that record to the matching unit 124 (S804). Note that the processes of S802 and S804 may be executed in reverse order or in parallel.

[0106] The collation unit 124 compares (or collates) the input character string with the conversion target character string in the conversion rule 201, and checks whether or not there is a character string in the input character string that matches the conversion target character string (S808).

[0107] If the result of the confirmation process is that the existence of a matching string is not confirmed (S808; NO), the conversion processing unit 102, for example, causes the matching unit 124 to read the next record in the conversion list 200 (S816) and re-executes the matching process of S806 using the conversion rule 201 of the next record.

[0108] This matching process is repeated by sequentially applying the next record in the conversion list 200 until a record of the conversion rule 201 that matches the input character string is found (until a YES determination is made in S808).

[0109] If the existence of a matching character string is confirmed in S808 (S808; YES), the conversion processing unit 102 converts the matching character string into a correct character string in the conversion unit 126 according to the conversion rule 201 regardless of whether the conversion is correct or incorrect (S810).

[0110] Then, the conversion processing unit 102 records the conversion processing result by the conversion unit 126 as a conversion history in the format 400 as shown in FIG. 4, for example, by the log output unit 128 (S812).

[0111] Thereafter, the conversion processing unit 102 checks, for example, whether there is a record of the conversion rule 201 that has not been applied in the conversion list 200, in other words, whether the record of the conversion rule 201 that was applied in S810 is the last record (S814).

[0112] If the result of the confirmation process indicates that there is a record of the conversion rule 201 that has not been applied (YES in S814), the conversion processing unit 102, for example, causes the matching unit 124 to read the next record in the conversion list 200 (S816) and re-executes the processes from S806 onwards.

[0113] The conversion processing unit 102 sequentially reads the next record in the conversion list 200 (S816) and repeats the processes from S806 onwards until there is no record in the conversion list 200 for which the conversion rule 201 has not been applied (until a NO determination is made in S814).

[0114] Here, in the conversion list 200, the records of the conversion rule 201 are arranged (sorted) in ascending order of the number of characters in the correct string as described above, so that by the repetitive process described above, the conversion rule 201 is applied to the input string in ascending order of the number of characters in the correct string.

[0115] If all conversion rules 201 in the conversion list 200 have been applied (NO in S814), the conversion processing unit 102 outputs log data 203 (see Figure 4) indicating the conversion history recorded for each execution of S812 to the memory unit 20 or the display unit 30 (S818).

[0116] FIG. 9 shows an example in which the character string "index spectrum" included in the input character string is converted sequentially by applying the conversion rules 201 in ascending order according to the above conversion process in the conversion processing unit 102.

[0117] As shown in FIG. 9, when the matching unit 124 applies the conversion rule 201-1 with the record number "0490068" to the input character string and performs matching, the substring "spectrum" of the character string "index spectrum" matches the character string to be converted, "[bebepepe]ku[tottor]ru."

[0118] Therefore, the substring "spectrum" is converted into the correct string "vector" by the conversion rule 201-1. Through this conversion, the pre-conversion string "index spectrum" becomes "index vector."

[0119] Next, in the matching unit 124, when the conversion rule 201-2 for the record number = "0590069" is applied to the input character string including the converted character string "index_svector" and a match is performed, it is found that "svector", which is a substring of the character string "index_svector", matches the character string to be converted "S[bebepepe]ku[totr]ru".

[0120] Therefore, the partial character string "specter" is converted into the correct character string "spectrum" by the conversion rule 201-2. Through this conversion, the character string before conversion "indexspecter" becomes "indexspectrum".

[0121] If there are no conversion rules 201 registered in the conversion list 200 for correct words with a larger number of characters other than the two conversion rules 201-1 and 201-2 described above, the conversion process ends and "index spectrum" becomes the final conversion process result.

[0122] Here, if "index vector" is a word with a correct string, the conversion rule 201 corresponding to the string "index spectrum", which is the final result of the conversion process, is not registered in the conversion list 200, and therefore the incorrect conversion will not be corrected.

[0123] Therefore, in this embodiment, a new conversion rule 201 corresponding to the character string for which such an incorrect conversion has been confirmed is generated by the conversion rule generation processing unit 104 in accordance with the generation processing example described below with reference to Figure 10, and added to the conversion list 200.

[0124] <Example of conversion rule generation process> 10 is a flowchart showing an example of a process for generating a conversion rule according to an embodiment. As illustrated in FIG. 10, the conversion rule generation processing unit 104, for example, in the character string generation unit 142, compares each of the correct character strings constituting the correct word 504 with the correct characters registered in the character recognition correction data 502, and identifies all correct characters that may be misrecognized and all misrecognized characters corresponding to the correct characters (S902).

[0125] The correct word 504 is created in advance, for example, by a user of the information processing device 1 checking the log data 203 described above, identifying a word (e.g., "index spectrum") that contains an incorrectly recognized character, and correcting the word to a correct character string.

[0126] Then, the character string generating unit 142 determines whether the correct character string contains one character that can be erroneously recognized, and whether there is one erroneously recognized character for that one character (in other words, the correct character) (S904).

[0127] If the result of the determination in S904 is "YES", the character string generation unit 142 generates a character string to be converted without using the regular expression [ ] (S906). For example, the character string generation unit 142 generates a character string to be converted by replacing the corresponding character of the correct character string with an incorrectly recognized character.

[0128] On the other hand, if the result of the determination in S904 is "NO", the character string generation unit 142 determines whether the number of regular expressions [ ] used in the character string to be converted is 1 (S908). For example, if the correct character string contains one character that may be misrecognized and there are multiple misrecognized characters for that one character (correct character), the correct character can be replaced with the regular expression [ ], and the number of regular expressions [ ] is 1 (YES in S908).

[0129] In this case, the string generation unit 142 generates a string to be converted by including multiple misrecognized characters but not the correct character in the square brackets [] of the regular expression, and replacing the correct character in the correct string with the regular expression [] (S910).

[0130] On the other hand, if there are multiple characters in the correct string that may be misrecognized, the number of regular expressions [ ] is determined to be multiple (NO) in S908, and the string generation unit 142 replaces each of the multiple characters with the regular expression [ ].

[0131] In this case, the string generation unit 142 generates a string to be converted by including a correct character and one or more misrecognized characters in each of the multiple regular expressions [ ] and replacing each of the correct characters that may be misrecognized in the correct string with the regular expression [ ] (S912).

[0132] The conversion target character string generated in character string generation unit 142 is passed to conversion rule generation unit 144 together with the correct answer character string, and conversion rule generation unit 144 generates conversion rule 201 by associating the conversion target character string with the correct answer character string (S914). The generated conversion rule 201 is passed from conversion rule generation unit 144 to management unit 146.

[0133] The management unit 146 adds, for example, a record number (the first two digits representing the number of characters in the correct string + a five-digit number obtained by incrementing the existing incremental value by one) and information indicating the number of characters in the correct string to the conversion rule 201, and adds this record to the conversion list 200 (S916).

[0134] When a record of a new conversion rule 201 is added to the conversion list 200, the management unit 146 may output log data related to the update of the conversion list 200 to the storage unit 20 or the display unit 30, for example, for the purpose of maintaining, operating, and managing the conversion list 200. This log data may have a format different from that of the log data 203 exemplified in FIG.

[0135] Fig. 11(a) shows an example of a record of the conversion rule 201 to be newly added to the conversion list 200. Fig. 11(a) shows an example of a record of the conversion rule 201 corresponding to the character string "index spectrum" that was erroneously converted in Fig. 9. The record number of this record is, for example, "1090071", which is a combination of the first two digits indicating the number of correct characters (=10 characters) and the subsequent five digits "90071".

[0136] Furthermore, the conversion target character string in the conversion rule 201 of this record is generated by the process of S912 because the seventh and ninth correct characters of the correct character string "index vector" may be misrecognized. For example, the character string "index[bebepepe]ku[tottor]ru" is generated as the conversion target character string by using the regular expression [] at each of the seventh and ninth character positions.

[0137] Then, the management unit 146 adds, for example, a record of the additional conversion rule 201 to the conversion list 200, and sorts the records of the conversion list 200 in ascending order using a record number based on the number of characters in the correct character string as a key (S918). As a result, the conversion rules 201 are rearranged in the conversion list 200 in ascending order of the number of correct characters. Note that Figure 11(b) shows an example of the result when the record of the conversion rule 201 exemplified in Figure 11(a) is added to the conversion list 200 exemplified in Figure 7 and sorted.

[0138] In this way, a new conversion rule 201 corresponding to the character string in which the erroneous conversion occurred is generated, and the conversion list 200 is updated. Therefore, as the conversion list 200 is updated, the probability of occurrence of a case in which the erroneous conversion cannot be completely corrected is adaptively reduced.

[0139] For example, the erroneously converted character string "index spectrum" in FIG. 9 is converted to the correct character string "index vector" as shown in FIG. 12 by applying the conversion rule 201 of the newly added record number "1090071" shown in FIG. 11, and the erroneous conversion is eliminated.

[0140] As described above, according to the above-described embodiment, the conversion rule 201 is applied in order from the character string with the fewest number of characters to the character string with the largest number of characters (i.e., ascending order), and conversion to a correct character string is executed sequentially. Therefore, even if an erroneous conversion occurs during the conversion process by the conversion processing unit 102, the probability that the erroneous conversion can be resolved is improved.

[0141] For example, the probability that a first character string will be converted into a first correct character string and then erroneously converted into a second correct character string that includes the first correct character string and has a larger number of characters can be reduced. This improves the accuracy of converting a misrecognized character string into a correct character string.

[0142] Furthermore, in the conversion list 200, the order of the conversion rules 201 is managed in ascending order of the number of characters in the correct answer string, which eliminates the need for, for example, the control or management of the application order of the conversion rules 201 by the conversion processing unit 102. Therefore, for example, processing efficiency is improved when mechanically executing conversion processing using the conversion list 200.

[0143] Furthermore, in the conversion list 200, the conversion rules 201 are managed by record numbers having a number format based on the number of characters in the correct string, which improves processing efficiency for managing the order of the conversion rules 201 in ascending order, for example.

[0144] Furthermore, when a new erroneous conversion is discovered, the conversion rule generation processing unit 104 generates a conversion rule 201 corresponding to the new erroneous conversion and adds it to the conversion list 200, making it possible to flexibly adapt to the new erroneous conversion.

[0145] <Modification> Next, modified examples of the above-described embodiment will be described with reference to Figures 13 to 15. In the modified examples, the configuration of the information processing device 1 may be the same as the configuration exemplified in Figure 2, and elements described with the same reference numerals as those already described may be understood to be the same as or similar to the already described elements unless otherwise specified.

[0146] Figures 13(a) and 13(b) are diagrams showing an example of a conversion list 200 according to a modified example. The conversion list 200 shown in Figures 13(a) and 13(b) differs from the conversion list 200 shown in Figures 7 and 11(b) in that its format includes a flag field 210 as an example of control information that controls (or indicates) whether or not to apply the conversion rule 201 for each record of the conversion rule 201. A flag indicating "0" or "1" is set in the flag field 210.

[0147] The flag setting can be performed by, for example, the conversion rule generation processing unit 104 (for example, the management unit 146) illustrated in Fig. 2. After the conversion rule generation processing illustrated in Fig. 10, the conversion rule generation processing unit 104 performs the following additional processing on each record from the first to the last of the conversion list 200 sorted in ascending order of the number of characters, thereby setting the flag on a record-by-record basis.

[0148] In Fig. 13(a), the character string "Vector" in bold indicates that it is included as part of the words "Bias Vector" and "Index Vector" in a character string longer than the character string in question. Similarly, in Fig. 13(b), the character string in bold indicates that it is included as part of another word in a character string longer than the character string in question.

[0149] <Record comparison processing> The conversion rule generation processing unit 104, for example, sequentially checks from the higher entries (or records) in the conversion list 200 whether a character string corresponding to a word in the conversion rule 201 of the entry of interest (hereinafter also referred to as the "target record") is included in the conversion target character string of the conversion rule in the entry lower than the entry of interest.

[0150] As a non-limiting example, the conversion rule generation processing unit 104 sequentially compares and checks whether the "conversion target string" in the target record is included in the "correct answer string" in the conversion rule 201 of a lower record whose first two digits representing the number of characters are greater than those of the target record.

[0151] For example, in Figure 13(a), a comparison is made to see if a string that matches any of the multiple string patterns represented by the conversion target string "[bebepepe]ku[totto]ru" in the target record with record number "0490068" is included in the "correct answer string" of a record lower than the target record.

[0152] In FIG. 13(a), the conversion target character string "vector" of record number "0490068" is included in the correct character string "bias vector" of record number "0890070" and the correct character string "index vector" of record number "1090071".

[0153] <Replacement process> If the comparison reveals that a lower-level record exists that includes the "conversion target string" in the target record as part of its "correct answer string," the conversion rule generation processing unit 104 replaces the part of the "conversion target string" in the lower-level record that corresponds to the "correct answer string" in the target record with the "correct answer string" in the target record.

[0154] For example, in FIG. 13(a), the "vector" portion in each of the conversion target character strings with record numbers "0890070" and "1090071" is replaced with the correct character string "vector" in the target record with record number "0490068" which has a smaller number of characters.

[0155] Therefore, the "subject string" for the correct string "bias vector" of record number "0890070" will have the same "bias vector" as the correct string. Similarly, the "subject string" for the correct string "index vector" of record number "1090071" will have the same "index vector" as the correct string.

[0156] As illustrated above, when the conversion rule generation processing unit 104 confirms that a character string corresponding to a word in the conversion rule 201 of the entry of interest in the conversion list 200 is included in the conversion target character string of a conversion rule in an entry lower than the entry of interest, the conversion rule generation processing unit 104 replaces the part of the conversion target character string corresponding to the word in the conversion rule 201 of the entry of interest with the word in the conversion rule 201 of the entry of interest.

[0157] <Flag setting process> The conversion rule generation processing unit 104 (e.g., management unit 146) checks whether the conversion target string of the conversion rule 201 in an entry lower than the entry of interest, which includes the word replaced in the above manner, is identical to the word associated with it in the conversion rule 201 of the lower entry.

[0158] If the result of the check is "same," conversion rule generation processing unit 104 records, for example, a flag indicating "same" = 1 in flag field 210 of the lower entry. On the other hand, if the result of the check is "different," conversion rule generation processing unit 104 records, for example, a flag indicating "not same" = 0 in flag field 210 of the lower entry. Note that "recording" a flag may be interpreted interchangeably as "setting" or "assigning" a flag.

[0159] For example, in Figure 13(a), the "correct string" and the "conversion target string" after replacement for record numbers "0890070" and "1090071" are the same string. Therefore, flag = 1 is set in the flag field 210 of each of the records with record numbers "0890070" and "1090071." Note that in Figure 13(a), there are no records with a number smaller than record number "0490068," so flag = 0 is set in the flag field 210 of the record with record number "0490068."

[0160] In this way, the flag indicating "same" = 1 is set when both the first and second conditions shown below associated with the "flag" are met.

[0161] First condition: The correct answer string in a lower record with a higher record number than the target record contains the string to be converted in the target record.

[0162] Second condition: A string obtained by replacing a part of a conversion target string in a lower record that corresponds to the correct string of the target record with the correct string of the target record is identical to the correct string of the lower record.

[0163] If the first condition is met but the second condition is not met, i.e., if the "correct string" in the lower record and the "string to be converted" after replacement are different strings, the conversion rule generation processing unit 104 sets a flag indicating "not identical" = 0 in the flag field 210 of the lower record.

[0164] For example, as shown in FIG. 13(b), assume a case where a record (record number "0590072") of the conversion rule 201 in which the correct character string is "spectrum" is added to the conversion list 200 illustrated in FIG. 13(a).

[0165] In this case, the conversion target string "Pektor" for the correct answer string "Vector" of record number "0490068" is included in the correct answer string "Spectrum" of the lower record number "0590072" which has a larger number of characters. Note that the conversion target string "Pektor" of record number "0490068" is one of the multiple string patterns represented by the conversion target string "[Bebepepe]ku[totr]ru".

[0166] Therefore, the "pektor" part in the conversion target string for the correct answer string "spectrum" of record number "0590072" is replaced with the correct answer string "vector" of record number "0490068" which has a smaller number of characters by the replacement process described above.

[0167] As a result, the conversion target character string for the correct answer character string "spectrum" of record number "0590072" is "spectrum". This conversion target character string "spectrum" is a character string different from the correct answer character string "spectrum". Therefore, the conversion rule generation processing unit 104 sets flag = 0 in the flag field 210 of the record with record number "0590072".

[0168] Furthermore, the conversion target character string "subject" of this record number "0590072" is included in the "bias vector" and "index vector", which are the correct character strings of the records with record numbers "0890070" and "1090071", respectively.

[0169] Therefore, the part corresponding to "spectrum" in the character string to be converted in each of these records is replaced with "spectrum", which is the correct character string for record number "0590072", by the replacement process described above.

[0170] As a result of this substitution, as shown in Figure 13(b), the conversion target character string for the correct character string "bias vector" in record number "0890070" becomes "bia spectrum." Similarly, the conversion target character string for the correct character string "index vector" in record number "1090071" becomes "index spectrum."

[0171] The correct string and the string to be converted are different strings in each of the records with record numbers "0890070" and "1090071." Therefore, the conversion rule generation processing unit 104 sets flag=0 in the flag field 210 of each of the records with record numbers "0890070" and "1090071."

[0172] An example of processing including the inter-record comparison processing, partial string replacement processing, and flag setting processing is shown as a flowchart in Fig. 14. The flowchart shown in Fig. 14 is executed in the conversion rule generation processing unit 104 (e.g., the management unit 146) after the flowchart shown in Fig. 10 is executed, for example.

[0173] As shown in FIG. 14, for example, the conversion rule generation processing unit 104 reads the first record of the conversion list 200 sorted in ascending order of the number of characters in the correct answer string as the "target record" from the storage unit 20 (S1402).

[0174] Then, the conversion rule generation processor 104 compares the conversion target character string of the "target record" with the correct answer character string of the lower "other record" that has a larger number of characters, and checks whether the conversion target character string of the "target record" is included in the correct answer character string of the "other record" (S1404). This check corresponds to checking whether the first condition described above is satisfied.

[0175] If the result of the confirmation indicates "YES", the conversion rule generation processing unit 104 replaces the part of the conversion target character string of the "other record" that corresponds to the correct character string of the "target record" with the correct character string of the "target record" (S1406).

[0176] Then, the conversion rule generation processing unit 104 checks whether the converted character string after the replacement is the same as the correct character string (S1410). This check corresponds to checking whether the second condition described above is satisfied.

[0177] If the result of the check is that they are the same (YES in S1410), conversion rule generation processing unit 104 sets flag = 1 in flag field 210 of the "other record" (S1412). If they are not the same (NO), conversion rule generation processing unit 104 sets flag = 0 in flag field 210 of the "other record" (S1414).

[0178] After setting the flag, the conversion rule generation processing unit 104 checks whether the "target record" is the last record in the conversion list 200 (S1416). If the "target record" is not the last record in the conversion list 200 (NO in S1416), the conversion rule generation processing unit 104 reads the next record in the conversion list 200 as the "target record" (S1418) and repeats the processing from S1404 onwards.

[0179] In addition, if it is determined in the processing of S1404 that the conversion target string of the "target record" is not included in the correct answer string of the "other records" (NO), the conversion rule generation processing unit 104 reads the next record in the conversion list 200 as the "target record" (S1418) and repeats the processing from S1404 onwards.

[0180] Then, when conversion rule generation processor 104 has completed setting the flag (S1412 or S1414) for the last record in conversion list 200 (YES in S1416), conversion rule generation processor 104 ends the process.

[0181] The flag set for each record of the conversion list 200 in this manner is used as an example of information for controlling whether or not to apply the conversion rule 201 in the conversion process (see, for example, FIG. 8) using the conversion list 200 described above. For example, flag = 0 indicates "apply," and flag = 1 indicates "not apply."

[0182] Therefore, in the application of the conversion rules 201 in ascending order, the conversion rules 201 of the records with flag = 0 are applied, and the conversion rules 201 of the records with flag = 1 are not applied (in other words, they are excluded or skipped from application).

[0183] In other words, as already mentioned, flag = 1 indicates that the string to be converted and the correct string are the same string in conversion rule 201, so by not applying conversion rule 201 with flag = 1, redundant (or unnecessary or wasteful) conversion of the same string into the same string is prevented.

[0184] Fig. 15 shows an example of a flowchart in which the above-described flag determination process is added to the conversion process shown in Fig. 8. As shown in Fig. 15, the flag determination process (S1505) according to this modification is performed, for example, after the record reading process (S804 and S816) and before the string matching process (S806).

[0185] Therefore, if the flag field 210 of the read record is flag=0 (YES in S1505), the conversion processing unit 102 determines to apply the conversion rule 201 of the record, and executes the processes from S806 onwards described in FIG.

[0186] On the other hand, if the flag field 210 has flag=1 (NO in S1505), the conversion processing unit 102 decides not to apply the conversion rule 201 of the record, and repeats reading the next record (S816) until a record with flag=0 is found.

[0187] As described above, according to this modification, the application of the conversion rule 201 that would cause redundant conversion due to the flag determination process is skipped, so that the efficiency of the conversion process can be improved.

[0188] <Additional information> The information processing device 1 illustrated in Fig. 2 may be configured with at least one computer. The computer may be a server computer or a client computer. The processing executed by the computer may be distributed and executed among multiple computers, including the computer. The multiple computers may be, for example, a group of cloud servers virtualized by virtualization technology.

[0189] 2, the above-described embodiment has been described with reference to a mode in which the conversion processing unit 102 and the conversion rule generation processing unit 104 are provided in the processing unit 10 of one information processing device 1. However, the present invention is not limited to such a mode, and the conversion processing unit 102 and the conversion rule generation processing unit 104 may be configured as separate information processing devices (e.g., computers), for example.

[0190] <Example of hardware configuration of information processing device 1> 16 is a diagram showing an example of the hardware configuration of a computer 1300 that functions as an information processing device 1 according to an embodiment. Physically, the computer 1300 may be configured to include, for example, a central processing unit (CPU) 1302, a storage device 1304, an input device 1306, and an output device 1308. These devices are connected to each other so as to be able to communicate with each other by a bus 1310 such as an address bus or a data bus.

[0191] The term "apparatus" may be interchangeably read as other terms such as circuit, device, unit, etc. The hardware configuration of computer 1300 may be configured to include one or more of the apparatuses shown in FIG. 16, or may be configured to exclude some of the apparatuses.

[0192] Each function of the computer 1300 is realized, for example, by the CPU 1302 reading and executing the program 205 or data used in conjunction with the execution of the program 205 from the storage device 1304. The data used in conjunction with the execution of the program 205 is, for example, the conversion list 200 having the data structure illustrated in Fig. 7. The CPU 1302 may be a single-core CPU or a multi-core CPU.

[0193] The program 205 is a program that, when read and executed by the CPU 1302, causes the computer 1300 to execute at least a part of the functions or processes of the information processing device 1 described in the above-described embodiment.

[0194] For example, the processing unit 10 including one or both of the conversion processing unit 102 and the conversion rule generation processing unit 104 illustrated in FIG. 2 can be realized by the CPU 1302 executing a program 205 that describes an instruction set that specifies the exemplary functions or processes described with reference to FIGS. 3 to 15.

[0195] The program 205 may be stored in advance in the storage device 1304, or may be received by the computer 1300 via a computer-readable storage medium or a network (telecommunications line) and stored in the storage device 1304. The program 205 may also be incorporated as part of the operating system (OS) of the computer 1300.

[0196] The storage device 1304 is an example of a computer-readable storage medium, and may correspond to the storage unit 20 illustrated in Fig. 2. For example, the storage device 1304 stores the conversion list 200, the log data 203, and the program 205 described above, and the CPU 1302 reads out at least one of the conversion list 200, the log data 203, and the program 205 as appropriate.

[0197] The storage device 1304 may include at least one of various storage media such as a read-only memory (ROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a random access memory (RAM), and a solid-state drive (SSD). These various storage media may be used, for example, as a working memory for the CPU 1302.

[0198] Additionally or alternatively, the storage device 1304 may be configured to include at least one of various storage media, such as an optical disk such as a CD-ROM (Compact Disc ROM), a hard disk drive (HDD), an SSD, a flexible disk, a magneto-optical disk (e.g., a compact disk, a Blu-ray (registered trademark) disc), and a flash memory.

[0199] The input device 1306 is an input device (for example, a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) that provides various input operation environments for the computer 1300 and accepts input to the computer 1300. In addition, the input device 1306 may include, for example, a device that accepts input by voice recognition or gesture operation.

[0200] The input device 1306 can input the character recognition correction data 502 and correct word 504 illustrated in FIG. 2 to the CPU 1302 functioning as the processing unit 10 (e.g., the conversion rule generation processing unit 104) via the memory device 1304 or without via the memory device 1304.

[0201] The output device 1308 is an output device (for example, a display corresponding to the display unit 30 illustrated in FIG. 2, a speaker, etc.) that performs output to an external device. The display is a device that displays the state of an input operation or a calculation result corresponding to the input operation, and may be, for example, a liquid crystal display or an organic EL (Electro Luminescence) display. Note that the input device 1306 and the output device 1308 may be integrated into one device, such as a touch panel.

[0202] The hardware configuration illustrated in FIG. 16 is merely an example and is not limited thereto. Part or all of the hardware configuration may be constructed in software using, for example, virtualization technology. Also, part or all of the devices in the hardware configuration may be located in different locations, and the devices may operate remotely in cooperation with each other via a network. Furthermore, the hardware configuration illustrated in FIG. 16 may be shared to operate part or all of multiple information processing systems.

[0203] Although the present disclosure has been described in detail above, it is clear to those skilled in the art that the spirit and scope of the present disclosure are not limited to the contents described throughout the present disclosure. The present disclosure can be implemented in modified and altered forms without departing from the spirit and scope of the present disclosure as defined by the claims. Therefore, the description of the present disclosure is intended for illustrative purposes only and does not have any limiting meaning on the spirit and scope of the present disclosure. [Explanation of symbols]

[0204] 1. Information processing equipment 10 Processing section 102 Conversion processing section 122 Acquisition Department 124 Matching Unit 126 Conversion Unit 128 Log output section 104 Conversion rule generation processing unit 142 String generator 144 Conversion rule generation unit 146 Management Department 20 Memory section 200 Conversion List 201-1~201-n Conversion Rules 210 Flags Field 203 Log Data 205 Programs 30 Display section 1300 Computer 1302 Central Processing Unit (CPU) 1304 Storage device 1306 Input Device 1308 Output Device 1310 Bus

Claims

1. a storage unit that stores a list of conversion rules that associates each of a plurality of words having different numbers of characters with a character string to be converted, for specifying a character string including a character that has been misrecognized as a conversion target; an information processing device comprising: a processing unit that compares the conversion rules with respect to a character string obtained by character recognition processing on a document in ascending order of the number of characters in the list, and sequentially converts character strings in the character string obtained by the character recognition processing that match the character string to be converted into corresponding words.

2. The information processing device according to claim 1 , wherein each of the conversion rules further comprises control information for controlling whether or not the conversion rule is applied in the processing unit.

3. a character string generation unit that generates the conversion target character string using character recognition correction data that associates correct characters that constitute a word with characters that may be erroneously recognized as the correct characters in a character recognition process for a document; a conversion rule generation unit that generates the list of conversion rules that associate the words with the character strings to be converted; The information processing device according to claim 1 , further comprising:

4. The character string generation unit When the number of correct characters that may be misrecognized in the word is one and there are a plurality of misrecognition candidate characters for the correct character, replacing the correct character with a regular expression that includes a plurality of the misrecognition candidate characters; When there are a plurality of correct characters that may be misrecognized in the word, each of the correct characters is replaced with a regular expression that includes a plurality of misrecognition candidate characters for the correct character and the correct character; and When there is one correct character that may be misrecognized in the word and there is one misrecognition candidate character for the correct character, the correct character is replaced with the misrecognition candidate character. The information processing apparatus according to claim 3 , wherein the conversion target character string is generated by executing at least one of the following.

5. The character string to be converted is: a character string in which one correct character that may be misrecognized in the word is replaced with a regular expression including a plurality of misrecognition candidate characters for the correct character; a character string in which each of a plurality of correct characters that may be erroneously recognized in the word is replaced with a regular expression including a plurality of erroneously recognized candidate characters for the correct character and the correct character; a character string in which one correct character that may be erroneously recognized in the word is replaced with one erroneously recognized candidate character for the correct character;

6. The conversion rules in the list are sorted in ascending order of the number of characters, The information processing device according to claim 1 , wherein the processing unit reads the conversion rules in the list in order from the top, and performs the matching in ascending order of the number of characters.

7. The processing unit 3. The information processing device according to claim 1, wherein the information processing device outputs log data that associates information indicating which conversion rule from the list was applied, a character string before conversion, a character string after conversion, a character string of a predetermined number of characters that includes the conversion target character string before conversion from among the character strings obtained by the character recognition processing, and a character string of the predetermined number of characters that includes the word after conversion from among the character strings obtained by the character recognition processing.

8. The information processing device reading out, from a storage unit that stores a list of conversion rules in which a conversion target character string is associated with each of a plurality of words having different numbers of characters, for specifying a character string including a misrecognized character as a conversion target, the conversion rules in ascending order of the number of characters; Sequentially comparing the conversion rules read in ascending order with a character string obtained by character recognition processing on a document, and sequentially converting character strings that match the conversion target character string in the character string obtained by the character recognition processing into corresponding words; An information processing method for performing processing including the steps of:

9. reading out, from a storage unit that stores a list of conversion rules in which a conversion target character string is associated with each of a plurality of words having different numbers of characters, for specifying a character string including a misrecognized character as a conversion target, the conversion rules in ascending order of the number of characters; Sequentially comparing the conversion rules read in ascending order with a character string obtained by character recognition processing on a document, and sequentially converting character strings that match the conversion target character string in the character string obtained by the character recognition processing into corresponding words; A program that causes an information processing device to execute the above.

10. A computer-readable storage medium storing the program according to claim 9.

11. a character string generation unit that, when a character is misrecognized in a character recognition process for a document, uses character recognition correction data that associates correct characters constituting a word with characters that may be misrecognized as the correct characters in the character recognition process, to identify a character string including the misrecognized character, and generates a conversion target character string based on the identified characters that may be misrecognized, in order to identify a character string including the misrecognized character as a conversion target; and a conversion rule generation unit that generates a conversion rule that associates the word with the character string to be converted.

12. The information processing device according to claim 11 , further comprising a conversion rule management unit that manages the conversion rules as a list in ascending order of the number of characters in the words.

13. the conversion rules are managed in a conversion rule management unit in units of records in which a record number based on the number of characters of the word, the word, and the character string to be converted are associated with each other; The information processing apparatus according to claim 11 , wherein the conversion rule management unit arranges the records in ascending order of the record numbers so as to be in ascending order of the number of characters in the words, and manages the records as a list.

14. The character string generation unit When the number of correct characters that may be misrecognized in the word is one and there are a plurality of misrecognition candidate characters for the correct character, replacing the correct character with a regular expression that includes a plurality of the misrecognition candidate characters; When there are a plurality of correct characters that may be misrecognized in the word, each of the correct characters is replaced with a regular expression that includes a plurality of misrecognition candidate characters for the correct character and the correct character; and When there is one correct character that may be misrecognized in the word and there is one misrecognition candidate character for the correct character, the correct character is replaced with the misrecognition candidate character.

14. The information processing apparatus according to claim 11, wherein the character string to be converted is generated by executing at least one of the following.

15. Each of the conversion rules further includes control information for controlling whether the conversion rule is applied; The conversion rule management unit Starting from the top entry of the list, it is checked in order whether a character string corresponding to a word in a conversion rule of the entry of interest is included in a character string to be converted in a conversion rule of an entry lower than the entry of interest; For the conversion target string corresponding to the entry confirmed to contain a string corresponding to the word, a portion corresponding to the word in the conversion rule of the entry of interest is replaced with the word in the conversion rule of the entry of interest; 14. An information processing device as described in claim 12 or 13, wherein, if the conversion target string of a conversion rule in an entry lower than the target entry, including the replaced word, is the same as the word in the lower entry, information indicating non-application of the conversion rule in the lower entry is recorded in the control information, and if they are different, information indicating application of the conversion rule in the lower entry is recorded in the control information.

16. The information processing device In character recognition processing of a document, when a character is misrecognized, in order to specify a character string including the misrecognized character, character recognition correction data that associates correct characters constituting a word with characters that may be misrecognized as the correct characters in the character recognition processing is used to specify characters that may be misrecognized for each correct character of the word, and based on the specified characters that may be misrecognized, a conversion target character string is generated to specify a character string including the misrecognized character as a conversion target; generating a conversion rule that associates the word with the character string to be converted; An information processing method for performing processing including the steps of:

17. The process reading out the conversion rules stored in a storage unit in ascending order of the number of characters in the words; Sequentially comparing the character string obtained by the character recognition process with the conversion rules read out in ascending order, and sequentially converting character strings in the character string obtained by the character recognition process that match the conversion target character string into corresponding words; The information processing method of claim 16 , further comprising:

18. In character recognition processing of a document, when a character is misrecognized, in order to specify a character string including the misrecognized character, character recognition correction data that associates correct characters that constitute a correct character string of the word with characters that may be misrecognized as the correct character in the character recognition processing is used to specify characters that may be misrecognized for each correct character of the word, and based on the specified characters that may be misrecognized, a conversion target character string is generated to specify a character string including the misrecognized character as a conversion target; generating a conversion rule that associates the word with the character string to be converted; A program that causes an information processing device to execute the above.

19. A computer-readable storage medium storing the program according to claim 18.

20. A data structure that defines a list of conversion rules used in an information processing device that includes a processing unit that processes character strings obtained by character recognition processing on a document, The list may include: A plurality of records of the conversion rule read by the processing unit, each of the plurality of records includes a record number based on the number of characters in a correct character string of a word in which a misrecognized character occurs in the character recognition process, the correct character string, and a conversion target character string for identifying a character string including a character in which a misrecognized character occurs in the character recognition process; The plurality of records are read by the processing unit according to the record number, and the processing unit executes a process of comparing the conversion rules with the character string obtained by the character recognition process in ascending order of the number of characters in the list, and sequentially converting character strings that match the character string to be converted into corresponding words.

21. 21. The data structure according to claim 20, wherein each of the records further includes control information for controlling whether or not the processing unit applies the conversion rule of the record.

22. The information processing device a first conversion for converting a first character string including characters misrecognized in a character recognition process for a document into a character string correct as a word of the first character string; a second conversion that converts a second character string that includes the character string obtained by the first conversion into a character string that is correct as a word of the second character string; An information processing method for performing processing including the steps of:

Images