JP2010140204A - Character recognition device, character recognition method, character recognition program, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a character recognition device, a character recognition method, a character recognition program and a recording medium for improving the precision of character recognition by preventing a recognized character string from being inappropriately replaced with a character string in a dictionary. <P>SOLUTION: A character recognition part 12 recognizes a character string included in a recognition object document 10. A similarity calculation part 15 calculates similarity between a character string in an itemized correction dictionary 14 and classified character strings classified by an item classification part 13, and determines a similar character string showing the maximum value of similarity. A threshold calculation part 17 calculates a threshold showing a degree of whether or not there is high possibility that the character string included in the recognition object document 10 exists in the itemized correction dictionary based on the update dates and recognition dates of the same classification items. A correction control part 18 compares the maximum value of similarity with the threshold, and when the maximum value of similarity is equal to or more than the threshold, the classified character strings are replaced with the similar character strings, and when the maximum value of similarity is less than the threshold, the classified character strings are not replaced with the similar character strings. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a character recognition device, a character recognition method, a character recognition program, and a recording medium that perform character recognition from an image including a character string, and more particularly, to a character that is recognized when character recognition is performed from an image including a character string. The present invention relates to a character recognition device, a character recognition method, a character recognition program, and a recording medium that improve character recognition accuracy by determining whether or not to correct a column and not performing unnecessary correction.

  Some electronic devices such as a personal computer (abbreviation “PC”) and a cellular phone are equipped with a character recognition device such as an optical character reader (abbreviation “OCR”). The character recognition device acquires image data of a document to be recognized by, for example, a camera or an image scanner. By performing pattern matching between the character string included in the image data and the character string in the character database stored in advance, the character string included in the image data is searched, and finally the recognition result in the form of the character code Output as.

  When character recognition is performed by a character recognition device, the output character recognition result may be incorrect due to, for example, the poor image state indicated by the acquired image data, or pattern matching with the character database not being performed properly. May occur. In order to correct an error in character recognition, the character recognition device prepares a dictionary containing a correct character string in advance, and if the recognized character string is a character string that does not exist in the dictionary, the recognition result Correct the string. The character recognition device compares the recognized character string with the character string in the dictionary to search for a similar character string, and replaces the recognized character string with the character string in the dictionary.

  In the character recognition device described in Patent Document 1, error patterns and correct answer patterns for long tones and hyphens are registered in a dictionary in advance, and when a character string that is a recognition result corresponds to an error pattern, A character string is corrected based on the correct answer pattern. When the character recognition device corrects the character string by mistake, it is possible to prevent the same erroneous correction from being performed again by the user newly registering a set of error patterns and correct patterns in the dictionary.

  In the character recognition device described in Patent Document 2, the recognized character string is cut out for each word and the spelling is checked. If the spelling check reveals that the word is incorrect, refer to the database in which the degree of misrecognition error is patterned on a character-by-character basis, correct the character in the incorrect word, and check the spelling again. If the correct word is confirmed, it is determined as a candidate word to be corrected. When there are a plurality of word candidates, the word to be finally corrected is determined by the user's selection.

  In the character recognition processing method described in Patent Literature 3, a word part is extracted from a recognized character string, and at least one word having a high degree of coincidence with the extracted word part is selected as a candidate from words in the word dictionary. . For each selected word, the frequency of occurrence in the document is calculated, and the candidate word is corrected to the word with the highest frequency value. In addition, information on other words related to each word can be registered in the word dictionary, and correction can be performed in consideration of not only the frequency of the word itself but also the frequency of the related word.

JP-A-4-268986 JP-A-5-298495 JP-A-9-44606

  In the character recognition device and the character recognition processing method described in Patent Literatures 1 to 3, when a correct character string exists in the dictionary, that is, a character string that can be recognized as being included in an image when viewed by a human being, If the character string in the dictionary matches, the character recognition result can be corrected correctly.

  However, if the correct character string does not exist in the dictionary, that is, if there is no character string in the dictionary that can be recognized as being included in the image when viewed by humans, character recognition is possible even if the dictionary is referenced. It is impossible to correct the result of. Also, in character recognition, even though the character string included in the image data is correctly recognized, look for a similar character string by referring to a dictionary where the correct character string does not exist, An incorrect recognition result may be output.

  Depending on the type of document to be recognized, a new character string that does not exist in the dictionary may be included. For example, when the document to be recognized is a business card, a character string representing an address, a zip code, a company name, a URL (Uniform Resource Locator), and the like is described. The address and postal code may be changed to a new character string by a merger of municipalities, and the company name and URL may be changed to a new character string by a merger of companies or a change of company name. It is also conceivable that new character strings such as new words and coined words appear in media such as newspapers and magazines.

  When any character string is registered in the dictionary, the amount of data increases, and searching and matching takes time, which causes a practical problem. The method of correcting a dictionary and registering a new character string may not always be effective because the word is not corrected as appropriate because new words and character strings may appear after the correction.

  As described above, when a certain period of time elapses after the dictionary is created, a new character string that does not exist in the dictionary may appear and be included in the document to be recognized. Since it is difficult to guess whether a correct character string exists in the dictionary, character recognition is performed correctly, and if the same character string as the recognized character string does not exist in the dictionary, it is recognized. Will be replaced with an inappropriate string in the dictionary.

  An object of the present invention is to prevent a character string to be recognized from being inappropriately replaced with a character string in a dictionary and improve the accuracy of character recognition, a character recognition method, a character recognition program, and It is to provide a recording medium.

The present invention relates to storage means for storing correction dictionary information representing a correction dictionary including a character string classified for each preset item in association with update date information representing a date when each item is updated, and
Recognizing means for recognizing a character string from image data representing an image including a character string to be recognized;
Classifying means for classifying the character string recognized by the recognition means into any of the preset items;
Among the correction dictionary items indicated by the correction dictionary information stored by the storage unit, for each character string of the same classification item that is the same item as the item into which the classified character string classified by the classification unit is classified, A similarity calculating means for calculating a similarity indicating a degree similar to the classified character string, and determining a character string indicating the maximum value among the calculated similarity values as a similar character string;
Based on a recognition date that is a date when the character string is recognized by the recognition unit and an update date indicated by update date information associated with the same classification item, the character string included in the image is stored in the correction dictionary. A threshold value calculation means for calculating a threshold value indicating a degree of whether or not there is a high possibility of existence;
Replacement means for controlling whether or not to replace the classified character string with the similar character string according to the maximum value of the similarity calculated by the similarity calculation means and the threshold value calculated by the threshold value calculation means Is a character recognition apparatus characterized by including.

The present invention also provides a computer,
Storage means for storing correction dictionary information representing a correction dictionary including character strings classified for each preset item in association with update date information representing a date when each item was updated;
Recognizing means for recognizing a character string from image data representing an image including a character string to be recognized;
Classifying means for classifying the character string recognized by the recognition means into any of the preset items;
Among the correction dictionary items indicated by the correction dictionary information stored by the storage unit, for each character string of the same classification item that is the same item as the item into which the classified character string classified by the classification unit is classified, A similarity calculating means for calculating a similarity indicating a degree similar to the classified character string, and determining a character string indicating the maximum value among the calculated similarity values as a similar character string;
Based on a recognition date that is a date when the character string is recognized by the recognition unit and an update date indicated by update date information associated with the same classification item, the character string included in the image is stored in the correction dictionary. A threshold value calculation means for calculating a threshold value indicating a degree of whether or not there is a high possibility of existence;
Replacement means for controlling whether or not to replace the classified character string with the similar character string according to the maximum value of the similarity calculated by the similarity calculation means and the threshold value calculated by the threshold value calculation means Is a character recognition program for functioning as

  The present invention is also a computer-readable recording medium on which the character recognition program is recorded.

The present invention also includes a recognition step of recognizing a character string from image data representing an image including a character string to be recognized;
A classification step of classifying the character string recognized by the recognition step into any of preset items;
Correction dictionary information stored by storage means for storing correction dictionary information representing a correction dictionary including a character string classified for each preset item in association with update date information representing a date when each item was updated The degree of similarity to the classified character string is indicated for each character string of the same classification item that is the same item as the item into which the classified character string classified by the classification step is classified. A similarity calculation step of calculating a similarity and determining a character string indicating the maximum value among the calculated similarity values as a similar character string;
Based on the recognition date, which is the date when the character string is recognized by the recognition step, and the update date indicated by the update date information associated with the same classification item, the character string included in the image is stored in the correction dictionary. A threshold calculation step for calculating a threshold indicating a degree of whether or not there is a high possibility of existence;
A replacement step for controlling whether or not to replace the classified character string with the similar character string according to the maximum value of the similarity calculated in the similarity calculation step and the threshold value calculated in the threshold value calculation step The character recognition method characterized by including these.

  According to the present invention, the correction dictionary information representing the correction dictionary including the character string classified for each preset item is stored in association with the update date information indicating the date when each item was updated by the storage unit. The The character string is recognized from the image data representing the image including the character string to be recognized by the recognition means. The character string recognized by the recognition means is classified into one of the preset items by the classification means. Of the items in the correction dictionary indicated by the correction dictionary information stored in the storage unit by the similarity calculation unit, the same classification item that is the same item as the item into which the classified character string classified by the classification unit is classified For each character string, a similarity indicating a degree similar to the classified character string is calculated, and a character string indicating the maximum value among the calculated similarity values is determined as a similar character string. Based on a recognition date that is a date when the character string is recognized by the recognition unit by the threshold calculation unit and an update date indicated by the update date information associated with the same classification item, the character string included in the image is A threshold value is calculated that indicates the degree of whether or not there is a high probability that the correction dictionary exists. The classified character string is replaced with the similar character string by the replacing unit according to the maximum value of the similarity calculated by the similarity calculating unit and the threshold calculated by the threshold calculating unit.

  Therefore, from the recognized character string item and the update date of the correction dictionary, the degree of possibility that the character string included in the image is likely to exist in the correction dictionary is quantified as a threshold value. It is possible to control whether or not to replace the recognized character string with the character string in the dictionary according to the maximum value of the similarity and the threshold value. Therefore, it is possible to provide a character recognition device that prevents improper replacement of a recognized character string with a character string in a dictionary and improves the accuracy of character recognition.

  Further, according to the present invention, the computer recognizes the correction dictionary information indicating the correction dictionary including the character string classified for each preset item by the character recognition program, and the update date information indicating the date when each item is updated. A storage means for storing the information in association with a recognition means, a recognition means for recognizing a character string from image data representing an image including a character string to be recognized, and a character string recognized by the recognition means as the preset item. Among the items of the classification means for classifying the data and the correction dictionary information indicated by the correction dictionary information stored by the storage means, the same item as the item for which the classified character string classified by the classification means is classified For each character string of the same classification item, a similarity indicating the degree of similarity with the classified character string is calculated, and a character string indicating the maximum value among the calculated similarity values Based on the similarity calculation means determined as a similar character string, a recognition date that is a date on which the character string is recognized by the recognition means, and an update date indicated by update date information associated with the same classification item, the image A threshold value calculating means for calculating a threshold value indicating whether or not there is a high possibility that a character string included therein is present in the modified dictionary, and a maximum value of the similarity degree calculated by the similarity degree calculating means, In accordance with the threshold value calculated by the threshold value calculation means, it can function as a replacement means for controlling whether or not to replace the classified character string with the similar character string.

  Therefore, it is possible to provide a character recognition program that causes a computer to function so that the recognized character string can be prevented from being inappropriately replaced with a character string in the dictionary and the accuracy of character recognition can be improved.

  According to the invention, the character recognition program is recorded by a computer-readable recording medium. Therefore, the character recognition program can be provided as a computer-readable recording medium.

  According to the invention, in the recognition step, the character string is recognized from the image data representing the image including the character string to be recognized. In the classification step, the character string recognized in the recognition step is classified into one of preset items. In the similarity calculation step, storage means for storing correction dictionary information representing a correction dictionary including a character string classified for each preset item in association with update date information representing a date when each item is updated is stored. Among the correction dictionary items indicated by the stored correction dictionary information, for each character string of the same classification item that is the same item as the classification character string classified by the classification step, the classified character string The similarity indicating the degree of similarity is calculated, and the character string indicating the maximum value among the calculated similarity values is determined as the similar character string. In the threshold value calculating step, a character string included in the image is based on a recognition date that is a date when the character string is recognized by the recognition step and an update date indicated by update date information associated with the same classification item. A threshold value indicating a degree of whether or not there is a high possibility of existing in the correction dictionary is calculated. In the replacing step, whether or not to replace the classified character string with the similar character string according to the maximum value of the similarity calculated in the similarity calculating step and the threshold calculated in the threshold calculating step. Control.

  Therefore, from the recognized character string item and the update date of the correction dictionary, the degree of possibility that the character string included in the image is likely to exist in the correction dictionary is quantified as a threshold value. It is possible to control whether or not to replace the recognized character string with the character string in the dictionary according to the maximum value of the similarity and the threshold value. Therefore, it is possible to provide a character recognition method that prevents improper replacement of the recognized character string with the character string in the dictionary and improves the accuracy of character recognition.

  FIG. 1 is a block diagram showing an outline of a character recognition device 1 according to an embodiment of the present invention. The character recognition method according to the present invention is executed by the character recognition device 1. The character recognition device 1 includes an image input unit 11, a character recognition unit 12, an item classification unit 13, an item-specific correction dictionary unit 14, a similarity calculation unit 15, a date acquisition unit 16, a threshold calculation unit 17, a correction control unit 18, and items. It includes another modified dictionary update unit 21 and a threshold value calculation method update unit 23.

  The character recognition device 1 is constituted by a computer system including a storage device (not shown) such as a semiconductor memory and a hard disk device, a central processing unit (hereinafter referred to as “CPU”) not shown, and a camera or an image scanner (not shown). Is done. The CPU executes the control program stored in the storage device, thereby causing the image input unit 11, the character recognition unit 12, the item classification unit 13, the item-specific correction dictionary unit 14, the similarity calculation unit 15, the date acquisition unit 16, Each function of the threshold value calculation unit 17, the correction control unit 18, the item-specific correction dictionary update unit 21, and the threshold value calculation method update unit 23 is realized.

  The image input unit 11 obtains image data representing an image of the recognition target document 10 which is a document including a recognition target character string by controlling a camera or an image scanner. The character recognition unit 12 as a recognition unit recognizes a character string included in the image data acquired by the image input unit 11 by a known character recognition method. Character recognition methods include, for example, a method of recognizing character strings by performing pattern matching with characters in a character database stored in advance for character strings included in image data, not for words but for each character. There is. The character recognition device 1 may be configured not to include a camera or an image scanner. In this case, image data representing an image of the recognition target document 10 may be input to the character recognition unit 12 from the outside.

The item classification unit 13 as a classification unit classifies the character string recognized by the character recognition unit 12 into preset items according to keyword dictionary information 31 and rule information 32 described later. The preset items are, for example, “last name”, “first name”, “company name”, “title”, “zip code”, “address”, “phone number”, “FAX number”, “email address”, “URL (
Uniform Resource Locator) ”and“ Other ”. The item is the same as the item of the item-specific correction dictionary indicated by the item-specific correction dictionary information stored in the item-specific correction dictionary unit 14 described later. The character strings classified by the item classification unit 13 are hereinafter also referred to as “classified character strings”. The item classification unit 13 classifies all recognized character strings into items, and classifies character strings that do not correspond to any item into the item “others”.

  FIG. 2 is a diagram illustrating an example of a character string classification result 30 in which character strings are classified into items. The character string classification result 30 shows an example of a result of classifying each character string included in the business card by the item classification unit 13 when character recognition is performed using the business card as the recognition target document 10. For example, the character string “Yamada” is the item “last name”, the character string “Taro” is the item “first name”, the character string “△△△ company” is the item “company name”, and the character string “person in charge” is the item “ The character string “123-4567” is the item “zip code”, the character string “XX city XX city XX town 12-34” is the item “address”, and the character string “07xxx-000” -0000 "is the item" telephone number ", the character string" 07xxx-111-1111 "is the item" FAX number ", the character string" yamada@aaa.co.jp "is the item" mail address ", the character The column “http://aaaa.co.jp/” is classified into the item “URL”.

  Since the character string classification result 30 is an example, the items in which the recognized character strings are classified differ in the displayed items and character strings if the types of documents to be recognized are different. Even when character recognition is performed on a business card, the results differ from the items and character strings shown in the character string classification result 30 shown in FIG. 2 depending on the design of the business card, the image shooting state, or the character string recognition state. It may become. An item having no character string to be classified results in a character string classification result 30 such as not displaying both the item and the character string, or displaying only the item and not displaying the character string.

  FIG. 3 is a diagram illustrating an example of the keyword dictionary information 31. The keyword dictionary information 31 is information representing a keyword dictionary in which character strings serving as keywords are classified for each item. For example, a character string such as “Osaka Prefecture”, “Nara Prefecture”, and “Tenri City” is included in the item “address”, and a character string such as “http: //” and “www” is included in the item “URL”. Character strings such as “TEL” and “Telephone” in the “Number”, character strings such as “Yamada”, “Suzuki” and “Sato” in the “Name” item, “President” and “General Manager” in the “Title” item , Character strings such as “head of division”, “section manager” and “section chief” are shown. The item classification unit 13 examines whether or not the recognized character string matches the character string in the keyword dictionary, and classifies the recognized character string into each item. For example, when “Osaka Prefecture” is included in the recognized character string, the recognized character string “Osaka Prefecture” matches the character string “Osaka Prefecture” in the item “address” of the keyword dictionary information 31. The recognized character string “Osaka Prefecture” is classified into the item “address”.

  FIG. 4 is a diagram illustrating an example of the rule information 32 regarding the number arrangement. The rule information 32 is information representing a rule that determines which item the recognized character string is classified based on the arrangement or arrangement of the numbers of the recognized character string. The rule information 32 shown in FIG. 4 includes, for example, “(3 numbers) (hyphen) (4 numbers) (hyphen) (4 numbers)” in the item “mobile number” and “( (3 numbers) (hyphen) (3 characters) (hyphen) (4 characters) ”,“ (FAX) ”(if the recognized character string includes a keyword related to FAX),“ "Postal code" is "(3 numbers) (hyphen) (4 numbers)" and item "Address" is "(If it does not match the rules and consists of numbers and hyphens, it is regarded as the address of the address and classified as an address. Each rule is shown.

  The item classification unit 13 confirms whether or not the recognized character string meets any of the rules, and classifies the character string into each item. For example, the character string “123-4567” corresponds to the rule “(3 numeric characters) (hyphen) (4 numeric characters)” indicating the postal code of the rule information 32, and is therefore classified into the item “postal code”. The rule information 32 is an example, and various rules can be set, for example, when there is no hyphen in a telephone number, or when a part of a number is enclosed in parentheses.

  The item-specific correction dictionary unit 14 serving as a storage unit controls the storage device and the like, and each item includes item-specific correction dictionary information representing the item-specific correction dictionary including character strings classified for each preset item. It is stored in association with update date information representing the updated date. The item-specific correction dictionary is a correction dictionary. The items set in advance are the same as the items classified by the item classification unit 13 described above. The update date information is information representing a date when an item including each character string is first created or a date when a character string in the item is changed by the item-specific correction dictionary update unit 21 described later.

  FIG. 5 is a diagram illustrating an example of the item-specific correction dictionary information 33. The item-specific correction dictionary information 33 indicating the item-specific correction dictionary includes the address correction dictionary information 34 indicating the address correction dictionary, the zip code correction dictionary information 35 indicating the zip code correction dictionary, and the URL correction dictionary information indicating the URL correction dictionary. 36 etc. are comprised.

  For example, the address correction dictionary information 34 includes “Yamadacho”, “Nagatacho”, “Enomotocho”, “Yamaguchicho”, “Tenri City”, “Nara City”, “Bunkyo Ward”, It includes a character string such as “Chiyoda-ku” and an update date “2007/4/1”. The address correction dictionary includes position information indicating the position in the item of each character string, and the position information includes numbers such as 1, 2, 3,... With a colon in front of the character string. It is expressed so that each number is not duplicated in the address correction dictionary.

  The postal code correction dictionary information 35 includes “000-0000”, “000-0001”, “000-0002”, “000-0004”, “000-0015”, “000-” classified into the item “postal code”. A character string such as “0999” and “000-9999” and an update date “2008/5/1” are included. The postal code correction dictionary includes position information indicating the position within each character string item, and the position information is a number such as 1, 2, 3,... With a colon in front of the character string. It is used so that each number is not duplicated in the postal code correction dictionary.

  The URL correction dictionary information 36 includes “http://aaaa.co.jp”, “http://bbbb.co.jp”, “http://cccc.co.jp” classified into the item “URL”. , “Http://dddd.co.jp”, “http://eeee.co.jp”, “http://ffff.co.jp”, “http://gggg.co.jp”, etc. It includes a character string and an update date “2008/1/1”. The URL correction dictionary includes position information indicating the position within each character string item. For example, the position information is numbered 1, 2, 3,... With a colon in front of the character string. It is expressed and used so that each number does not overlap in the URL correction dictionary. As described above, in the item-specific correction dictionary, the update date can be managed for each item because the character strings are classified for each item.

  The similarity calculation unit 15 that is a similarity calculation means indicates the degree of similarity to the classified character string for each character string of items classified into the same item as the classified character string among the items of the item-specific correction dictionary. Calculate similarity. Of the items in the item-specific correction dictionary, the same item as the item into which the classified character string is classified is hereinafter referred to as “same classification item”. The similarity calculation unit 15 sequentially compares each character string of the same classification item with the classified character string, and calculates a similarity indicating the degree of similarity. As a method for evaluating the degree of matching between two character strings, there is a known method such as DP (Dynamic Programming) matching. The similarity is calculated from the difference in the number of matching characters and the length of the character string, and the character string indicating the maximum value is searched from the calculated similarity dictionary and determined. A character string having a maximum similarity value is hereinafter referred to as a “similar character string”. Since the item-specific correction dictionary classifies each character string for each item, a similar character string can be determined by searching only the character string in the same classification item.

The date acquisition unit 16 measures a date and time (not shown) such as an RTC (Real Time).
Clock) or a receiving device (not shown) that receives radio waves for a radio clock, and the like, acquires a recognition date that is a date when the character recognition unit 12 recognized the character string from the image data, and sends it to the threshold calculation unit 17 send. The same recognition date is used for each classified character string classified by the item classification unit 13. The date acquisition unit 16 is configured to acquire the date when the character recognition unit 12 recognized the character string from the image data, but may be configured to acquire not only the recognized date but also the recognized time.

  The threshold calculation unit 17 serving as a threshold calculation unit receives the recognition date from the date acquisition unit 16, receives the update date of the same classification item from the item-specific correction dictionary unit 14, and is a post-update period that is a period from the update date to the recognition date Is calculated. A calculation formula for calculating a threshold value for each item is set in advance, and the threshold value of each classified character string is calculated according to the calculation formula based on the post-update period and sent to the correction control unit 18.

  The threshold indicates the degree of whether or not there is a high possibility that the character string included in the recognition target document 10 exists in the item-specific correction dictionary, and the corrected character string is converted into a similar character string by the correction control unit 18 described later. This is a value used as a reference for determining whether or not to replace, and the larger the threshold value, the more difficult the replacement is performed. Even if the post-update period is the same, the threshold varies depending on the type of item. For example, the item “URL” or the like is included in the recognition target document 10 if the period in which the contents of the item-specific correction dictionary are valid is relatively short, that is, the post-update period is long, because the character string is frequently changed. It is less likely that the character string to be present in the item-specific correction dictionary. Therefore, the threshold value calculation unit 17 sets a calculation formula so that a larger threshold value is obtained when the post-update period becomes longer, thereby making it difficult to replace the classified character string. In addition, since the frequency of changing the character string of the item “last name” is low, the period in which the content of the item-specific correction dictionary is valid is relatively long, that is, even if the post-update period is long, There is a high possibility that the character string included in is in the item-specific correction dictionary. Therefore, the threshold value calculation unit 17 sets a calculation formula so that a small threshold value is obtained even if the post-update period is long, so that the classified character string can be easily replaced.

  The threshold calculation unit 17 sets a calculation formula with different formulas or parameters for each item, and inputs the value of the updated period into the calculation formula to calculate the threshold. The calculation formula set for each item is set differently depending on the characteristics of the item, for example, a character string is easily added or changed. In the case of an item where the character string contained in the document to be recognized changes frequently and new contents often appear, in order to increase the threshold value, the calculation formula is made a quadratic function, for example, and the value of the post-update period is set. Input to quadratic function. In the case of an item in which the character string included in the document to be recognized does not change so frequently, in order to reduce the threshold value, for example, the calculation formula is a linear function and the value of the post-update period is input to the linear function. For example, when the calculation formula is a linear function, a threshold corresponding to the type of item can be calculated by setting a coefficient representing a slope or a constant representing an intercept for each item.

  FIG. 6 is a diagram for explaining a process of calculating a threshold value by the threshold value calculation unit 17. In the first step, the threshold calculation unit 17 receives the recognition date from the date acquisition unit 16. When the character recognition unit 12 performs character recognition, the date acquisition unit 16 acquires the recognition date and sends it to the threshold value calculation unit 17. For the character strings recognized at the same time, the same recognition date, for example, September 1, 2008, is used regardless of the item of the classified character string.

  In the second step, the threshold value calculation unit 17 receives update date information representing the update date of the same classification item from the item-specific correction dictionary unit 14. For example, the update date of the item “address” is April 1, 2008, the update date of the item “zip code” is January 1, 2008, and the update date of the item “URL” is July 1, 2008. The update date of the item “telephone number” is February 1, 2008.

  In the third step, the threshold value calculation unit 17 calculates a post-update period for each classified character string from the recognition date and the update date. For example, the post-update period of the item “address” is 5 months, the post-update period of the item “zip code” is 8 months, the post-update period of the item “URL” is 2 months, and the item “phone number” The period after renewal is 7 months.

  In the fourth step, the threshold calculation unit 17 calculates a threshold for each item. A function that is a calculation formula for calculating a threshold value for each item is set in advance, and the value of the updated period is input to the function to calculate each threshold value. For example, the calculation formula of the item “address” is the function a, the calculation formula of the item “zip code” is the function b, the calculation formula of the item “URL” is the function c, and the calculation formula of the item “phone number” is Function d.

  In the fifth step, the threshold value calculation unit 17 sends the respective threshold values calculated in the fourth step to the correction control unit 18. For example, the item “address” is the threshold A, the item “zip code” is the threshold B, the item “URL” is the threshold C, and the item “telephone number” is the threshold D. Each process shown in FIG. 6 is an example, and actually threshold values are calculated for all items of the classified character string.

  The correction control unit 18 serving as a replacement unit compares the similarity calculated by the similarity calculation unit 15 with the threshold calculated by the threshold calculation unit 17, and converts the classified character strings into similarities in the item-specific correction dictionary. Whether or not to replace with a character string is controlled, and the result of character recognition is output as a recognition result 19 in the form of a character code. When the threshold value calculation unit 17 calculates a small threshold value because it is highly possible that a character string included in the recognition target document 10 exists in the item-specific correction dictionary, the recognized character string is easily replaced with a similar character string. . When the threshold value calculation unit 17 calculates a large threshold value because there is a high possibility that the character string included in the recognition target document 10 exists in the item-specific correction dictionary, the recognized character string is easily replaced with a similar character string. .

  Three methods for controlling whether the correction control unit 18 replaces the classified character string with a similar character string will be described. In the first replacement control method, when the similarity value is greater than or equal to the threshold value, the recognized character string is replaced with the similar character string and output. When the similarity value is less than the threshold value, the recognized character string is replaced with the similar character string. Control to output without any problem.

  The second replacement control method calculates a difference between the maximum value of the similarity and the threshold, compares the calculated difference value with a predetermined difference value, and if the difference value is equal to or larger than a predetermined difference value, When the recognized character string is replaced with a similar character string and output, and the difference value is less than a predetermined difference value, control is performed so that the classified character string is output as it is without replacing the recognized character string with the similar character string. For example, a value that is 100 larger than the threshold value is used as the first reference value, and replacement is performed when the maximum similarity value is equal to or greater than the first reference value, or a value that is 50% smaller than the threshold value is used as the second reference value. When the maximum value of the degree is greater than or equal to the second reference value, control such as replacement is possible. By setting the predetermined difference value to 0, control similar to the first replacement control method can be performed.

  The third replacement control method calculates the ratio between the maximum value of the similarity and the threshold value, compares the calculated ratio with a predetermined ratio, and if the ratio is equal to or higher than the predetermined ratio, the recognized character string is determined as a similar character. If the ratio is less than a predetermined ratio, the classified character string is output as it is without replacing the recognized character string with a similar character string. The ratio can be expressed by, for example, “maximum value of similarity / threshold value × 100 (%)”. By arbitrarily setting a predetermined ratio, for example, it is possible to perform control such as replacement when the maximum value of similarity is 95% or more of the threshold value. By setting the predetermined ratio to 100%, the same control as the first replacement control method can be performed.

  The item-specific correction dictionary updating unit 21 serving as a correction dictionary updating unit obtains item-specific correction dictionary update information 20 for updating the item-specific correction dictionary stored in the item-specific correction dictionary unit 14 from the outside, and item-specific correction dictionary update unit 21. Based on the correction dictionary update information 20, the dictionary information and date information stored in the item-specific correction dictionary unit 14 are changed.

  The item-specific correction dictionary updating unit 21 controls an information processing apparatus (not shown) that can acquire information from a recording medium such as a compact disk or a semiconductor memory, and is recorded on a recording medium such as a compact disk or a semiconductor memory. The item-specific corrected dictionary update information 20 is acquired from the outside. Further, by providing the character recognition device 1 with a communication unit (not shown) that can communicate with the outside via a communication network or the like, the item-specific correction dictionary update unit 21 acquires the item-specific correction dictionary update information 20 via the communication unit. You can also

  FIG. 7 is a diagram illustrating an example of the item-specific correction dictionary update information 20. The item-specific correction dictionary update information 20 includes item information representing an item to be updated, position information representing a position in the item of the character string to be updated, and character string information representing the updated character string. . The item information indicates which item of the item-specific correction dictionary unit 14 is to be changed. The position information indicates where the character string to be changed is in the item. The item information and the position information are indexes indicating which character string is to be changed.

  In the item-specific corrected dictionary update information 20 shown in FIG. 7, the number “4” that is position information and the content “Yamadacho” that is the updated character string are shown for the item “address”, and the item “zip code”. The number “7” as the position information and the content “000-9991” as the updated character string are shown for the item “URL”, the number “6” as the position information for the item “URL” and the content “as the character string after the update” http://ffff.com ”. For example, when the item-specific correction dictionary update information 20 is the item “address”, the position information “4”, and the character string “Yamadacho”, the character string of number 4 of the address item in the item-specific correction dictionary is changed to Yamadacho. It shows that. The item-specific correction dictionary update information 20 includes only information related to a character string to be changed, and does not include information related to a character string that does not need to be changed.

  The threshold value calculation method updating unit 23 that is a threshold value calculation method update unit obtains from the outside threshold value calculation method update information 22 that is second update information for updating calculation formula information that represents a calculation formula for calculating a threshold value. The threshold calculation unit 17 stores calculation formula information representing a calculation formula for calculating a threshold for each preset item. The threshold calculation method update unit 23 changes the calculation formula information stored in the threshold calculation unit 17 based on the threshold calculation method update information 22.

  The threshold value calculation method updating unit 23 controls an information processing apparatus (not shown) that can acquire information from a recording medium such as a compact disk or a semiconductor memory, and is recorded on a recording medium such as a compact disk or a semiconductor memory. The threshold calculation method update information 22 is acquired from the outside. Further, by providing the character recognition device 1 with a communication unit (not shown) that can communicate with the outside via a communication network or the like, the threshold value calculation method update unit 23 may acquire the threshold value calculation method update information 22 via the communication unit. it can. Furthermore, the threshold value calculation method updating unit 23 can change the calculation formula when the character recognition device 1 is provided with an input unit (not shown) such as a keyboard and the user inputs a function or a parameter included in the function from the input unit.

  FIG. 8 is a diagram illustrating an example of the threshold value calculation method update information 22. To change the calculation formula for calculating the threshold, change the function that is the calculation formula, such as changing the quadratic function to a linear function, and change the coefficient indicating the slope of the linear function or the constant indicating the intercept, for example. For example, there is a method of changing parameters included in the function. The threshold value calculation method update information 22 includes item information indicating items to be updated and information indicating the updated calculation formulas or parameters included in the calculation formulas. The calculation formula for calculating the threshold is set in advance for each item, and the item information is an index indicating which item the calculation formula is to be changed.

  In the threshold value calculation method update information 22 shown in FIG. 8, the function “f_url (x) = ax ^ 2 + b” is indicated in the item “URL”, and the parameters “c, d” are indicated in the item “address”. . The threshold value calculation method update information 22 indicates that, for example, for the item “URL”, the calculation formula for calculating the threshold value is changed to the function “f_url (x) = ax ^ 2 + b”. A value for the post-update period is input to the variable x. The threshold calculation method update information 22 indicates that only the parameter included in the function is changed to “c, d” for the item “address”. By changing only the parameter, for example, only the coefficient of the variable can be changed without changing the function itself. The threshold value calculation method update information 22 is composed only of information relating to an item whose calculation formula is to be changed, and does not include information relating to an item which does not need to be changed.

  FIG. 9 is a flowchart showing processing of the character recognition method. When the character recognition device 1 performs character recognition of the recognition target document 10 including the character string to be recognized, the process proceeds to step S1 to start processing. In step S2, the image input unit 11 acquires image data representing an image of the recognition target document 10 using, for example, a camera or an image scanner.

  In step S3, which is a recognition step, the character recognition unit 12 recognizes a character string included in the image data. In step S4, which is a classification step, the item classification unit 13 classifies the character string recognized by the character recognition unit 12 for each item set in advance.

  In step S <b> 5, the date acquisition unit 16 acquires a recognition date that is a date on which character recognition is performed by the character recognition unit 12, and sends the recognition date to the threshold value calculation unit 17. In step S6, the correction control unit 18 determines whether or not replacement of all the classified character strings to be classified has been determined in step S11 described later. If the determination of whether or not replacement is possible has been performed for all classified character strings, the process proceeds to step S13, and if the determination of whether or not replacement is possible has not been performed for all classified character strings, the process proceeds to step S7.

  In step S <b> 7, the similarity calculation unit 15 calculates the similarity with the classified character strings for all character strings in the same item of the item-specific correction dictionary unit 14. The similarity is calculated only for the same classification item among the items in the item-specific correction dictionary. In step S8, the similarity calculation unit 15 determines a maximum value among the calculated similarity values and a similar character string that is a character string indicating the maximum value of the similarity. Steps S7 and S8 are similarity calculation steps.

  In step S9, the threshold value calculation unit 17 calculates the post-update period from the update date and the recognition date of the same classification item. In step S10, the threshold value calculation unit 17 calculates a threshold value from the item of the classified character string and the updated period. The threshold value calculation unit 17 calculates the threshold value by inputting the value of the updated period to a function set in advance for each item. Steps S9 and S10 are threshold value calculation steps.

  In step S11, the correction control unit 18 controls whether or not to replace the classified character string with a similar character string according to the maximum value of the similarity and the threshold value. Specifically, the control is performed by the first replacement control method, the second replacement control method, or the third replacement control method described above. If it is determined that replacement is to be performed, the process proceeds to step S12. If it is determined that replacement is not to be performed, the process proceeds to step S6. In step S12, the correction control unit 18 corrects the classified character string by replacing the classified character string with a similar character string, and proceeds to step S6. Steps S11 and S12 are replacement steps. Steps S7 to S12 are processes performed for each classified character string.

  In step S13, after it is determined in step S6 that all the classified character strings have been replaced or not, the correction control unit 18 outputs the result of character recognition, and the process proceeds to step S14 and the process ends. .

  When the character recognition device 1 is configured not to include a camera or an image scanner, step S2 can be omitted. In that case, in step S3, the character recognition unit 12 may acquire image data representing an image of the recognition target document 10 from the outside and recognize a character string included in the image data.

  The control program stored in the storage device of the character recognition device 1 includes a computer, a recognition unit that is the character recognition unit 12, a classification unit that is the item classification unit 13, a similarity calculation unit that is the similarity calculation unit 15, and a threshold calculation. This is a character recognition program for causing each unit to function as a threshold calculation unit that is a unit 17 and a replacement unit that is a correction control unit 18. Therefore, the character recognition program can be provided as a program for causing a computer to function as each unit of the character recognition device 1.

  In the character recognition device 1, the character recognition program is stored in a storage device such as a semiconductor memory or a hard disk device, but is not limited to these storage devices, and is recorded in a computer-readable recording medium. May be. The recording medium may be a recording medium that can be read by providing a program reading device as an external storage device (not shown) and inserting the recording medium therein, or may be a storage device of another device. .

  Any recording medium may be used as long as the stored character recognition program is accessed from a computer and executed. Alternatively, in any recording medium, the character recognition program is read out, the read character recognition program is stored in the program storage area of the storage device, and the character recognition program is executed. Good.

The recording medium configured to be separable from the main body is, for example, a tape recording medium such as a magnetic tape / cassette tape, a magnetic disk such as a flexible disk / hard disk, a CD-ROM (Compact Disk Read Only Memory) / MO (Magneto Optical). disk) / MD (
Disc recording media of optical discs such as Mini Disc) / DVD (Digital Versatile Disk) / Blu-ray Disc, IC (Integrated Circuit) cards including memory cards / card recording media such as optical cards, or mask ROM / EPROM ( Erasable
Programmable Read Only Memory / EEPROM (Electrically Erasable
Programmable Read Only Memory) or a recording medium that carries a fixed program including a semiconductor memory such as a flash ROM. A character recognition program for causing a computer to function as each unit of the character recognition device 1 can be recorded on the recording medium or the like. Therefore, the character recognition program can be provided as a computer-readable recording medium.

  Even if the character recognition program is not recorded on a computer-readable recording medium, the character recognition program can be downloaded from another device via a communication network and stored in the program storage area of the character recognition device 1. . Therefore, since the character recognition program can be transmitted by a medium having a function of transmitting information such as a communication network, it is provided as an information transmission medium for transmitting a character recognition program for functioning as each unit of the character recognition device 1. You can also

  As described above, the item-specific correction dictionary information representing the item-specific correction dictionary including the character string classified for each item set in advance by the item-specific correction dictionary unit 14 represents the date when each item was updated. Stored in association with information. The character recognition unit 12 recognizes a character string from image data representing an image including the character string to be recognized. The item classification unit 13 classifies the character string recognized by the character recognition unit 12 into any of preset items. Items by which the classified character string classified by the item classification unit 13 is classified by the similarity calculation unit 15 among the items of the item-specific correction dictionary indicated by the item-specific correction dictionary information stored by the item-specific correction dictionary unit 14. For each character string of the same classification item that is the same item as, a similarity indicating the degree similar to the classified character string is calculated, and the character string indicating the maximum value among the calculated similarity values is used as a similar character string It is determined. Included in the recognition target document 10 based on the recognition date, which is the date on which the character recognition unit 12 recognizes the character string, and the update date indicated by the update date information associated with the same classification item by the threshold calculation unit 17. A threshold value indicating the degree of whether or not there is a high possibility that the character string exists in the item-specific correction dictionary is calculated. The correction control unit 18 replaces the classified character string with the similar character string according to the maximum value of the similarity calculated by the similarity calculation unit 15 and the threshold calculated by the threshold calculation unit 17.

  Therefore, the threshold value is a degree of whether or not there is a high possibility that the character string included in the recognition target document 10 exists in the item-specific correction dictionary from the recognized character string item and the update date of the item-specific correction dictionary. It is possible to control whether to replace the recognized character string with the character string in the dictionary according to the maximum value of the similarity and the threshold value. Therefore, it is possible to provide a character recognition device that prevents improper replacement of a recognized character string with a character string in a dictionary and improves the accuracy of character recognition.

  Furthermore, the threshold value calculation unit 17 calculates a threshold value according to the post-update period determined by the certification date and the update date according to a calculation formula set in advance for each item, thereby depending on the item type and the post-update period. Since it is possible to control whether or not the replacement is performed by calculating the threshold value for each item, it is possible to provide a character recognition device that further improves the accuracy of character recognition. In addition, it is possible to set whether or not replacement is easily performed for each character string, not for the entire recognized character string.

  Furthermore, when the maximum value of the similarity is equal to or greater than the threshold, the correction control unit 18 replaces the classified character string with a similar character string. When the maximum value of the similarity is less than the threshold, the classified character string is replaced with the similar character string. By not being replaced with, it is possible to appropriately control whether or not to perform the replacement according to the maximum value of the similarity and the threshold value.

  Further, the correction control unit 18 calculates a difference between the maximum value of the similarity and the threshold value, and compares the calculated difference value with a predetermined difference value. If the difference value is greater than or equal to the predetermined difference value, the classified character string is replaced with a similar character string, and if the difference value is less than the predetermined difference value, the classified character string is not replaced with a similar character string. Whether or not the recognized character string is easily replaced can be set uniformly instead of for each item.

  Further, the correction control unit 18 calculates a ratio between the maximum value of the similarity and the threshold value, and compares the calculated ratio with a predetermined ratio. When the ratio is greater than or equal to the predetermined ratio, the recognized character string is replaced with a similar character string, and when the ratio is less than the predetermined ratio, the recognized character string is not replaced with the similar character string, so that the recognized character string is replaced. It is possible to uniformly set whether or not it is easy to be performed, not for each item.

  Further, the item-specific correction dictionary update unit 21 obtains item-specific correction dictionary update information 20 for updating the item-specific correction dictionary from the outside, and the item-specific correction dictionary update information 20 stores the item-specific correction dictionary update information 20. By updating the stored item-specific correction dictionary information and update date information, the item-specific correction dictionary managed for each item is updated, and the old character string in the item-specific correction dictionary is replaced with a new character string. Therefore, the accuracy of character recognition can be improved.

  Further, the item-specific correction dictionary update information 20 that is update information includes item information representing an item to be updated, position information representing a position in the item of the character string to be updated, and a character string representing a character string after the update. By including the information, the character string in the item-specific correction dictionary can be appropriately changed to a new character string using the item information and the position information as indexes.

  Further, the threshold calculation unit 17 stores calculation formula information representing a calculation formula for calculating the threshold for each item. The threshold calculation method update unit 23 obtains from the outside threshold calculation method update information 22 as second update information for updating the calculation formula information, and the calculation formula information is changed based on the threshold calculation method update information 22. Thus, since the calculation formula for calculating the threshold can be changed for each item, it is possible to arbitrarily adjust for each item whether or not the character string is easily replaced.

  Furthermore, the threshold value calculation method update information 22 that is the second update information includes item information that represents an item to be updated, and information that represents a calculation formula after update or a parameter included in the calculation formula, Using the item information as an index, the calculation formula for calculating the threshold value for each item can be appropriately changed.

  Further, by the character recognition program, the computer is used to update the item-specific correction dictionary information representing the item-specific correction dictionary including the character strings classified for each preset item, and the update date information representing the date when each item was updated. An item-specific correction dictionary unit 14 to be stored in association with each other, a character recognition unit 12 that recognizes a character string from image data representing an image including a character string to be recognized, and a character string recognized by the character recognition unit 12 Among the items of the item classification dictionary 13 indicated by the item classification dictionary 13 stored in the item classification unit 13 and the item-specific correction dictionary information stored in the item-specific correction dictionary unit 14, the item classification unit 13 classifies the item into one of the preset items. For each character string of the same classification item that is the same item as the classified character string to be classified, a similarity indicating the degree of similarity with the classified character string is calculated and calculated The similarity calculation unit 15 that determines a character string indicating the maximum value among the similarity values obtained as a similar character string, the recognition date that is the date on which the character string is recognized by the character recognition unit 12, and the same classification item Based on the update date indicated by the associated update date information, a threshold calculation for calculating a threshold indicating whether or not there is a high possibility that the character string included in the recognition target document 10 exists in the item-specific correction dictionary. Whether or not to replace the classified character string with the similar character string is controlled according to the maximum value of the similarity calculated by the unit 17 and the similarity calculating unit 15 and the threshold calculated by the threshold calculating unit 17. It can function as the correction control unit 18.

  Therefore, it is possible to provide a character recognition program that causes a computer to function so that the recognized character string can be prevented from being inappropriately replaced with a character string in the dictionary and the accuracy of character recognition can be improved.

  Furthermore, the character recognition program can be provided as a computer-readable recording medium.

  In step S3, which is a recognition step, a character string is recognized from image data representing an image including the character string to be recognized. In step S4, which is a classification step, the character string recognized in the recognition step is classified into one of preset items. In steps S7 and S8, which are similarity calculation steps, the item-specific correction dictionary information representing the item-specific correction dictionary including the character string classified for each preset item is used as the update date representing the date each item was updated. Of the items of the item-specific correction dictionary indicated by the item-specific correction dictionary information stored by the item-specific correction dictionary unit 14 stored in association with the information, the same item as the item into which the classified character string classified by the classification step is classified For each character string of the same classification item, the similarity indicating the degree of similarity with the classified character string is calculated, and the character string indicating the maximum value among the calculated similarity values is determined as the similar character string. In steps S9 and S10, which are threshold calculation steps, in the recognition target document 10, based on the recognition date, which is the date when the character string is recognized by the recognition step, and the update date indicated by the update date information associated with the same classification item. A threshold value indicating the degree of whether or not there is a high possibility that the character string included in the item exists in the item-specific correction dictionary is calculated. In steps S11 and S12, which are replacement steps, whether or not the classified character string is replaced with a similar character string according to the maximum value of the similarity calculated in the similarity calculation step and the threshold calculated in the threshold calculation step. To control.

  Therefore, the threshold value is a degree of whether or not there is a high possibility that the character string included in the recognition target document 10 exists in the item-specific correction dictionary from the recognized character string item and the update date of the item-specific correction dictionary. It is possible to control whether to replace the recognized character string with the character string in the dictionary according to the maximum value of the similarity and the threshold value. Therefore, it is possible to provide a character recognition method that prevents improper replacement of the recognized character string with the character string in the dictionary and improves the accuracy of character recognition.

It is a block diagram which shows the outline of the character recognition apparatus 1 which is one Embodiment of this invention. It is a figure which shows an example of the character string classification result 30 in which the character string was classified into each item. It is a figure which shows an example of the keyword dictionary information 31. FIG. It is a figure which shows an example of the rule information 32 regarding a number arrangement | positioning. It is a figure which shows an example of the correction dictionary information 33 classified by item. It is a figure for demonstrating the process of the process which calculates a threshold value by the threshold value calculation part. It is a figure which shows an example of the correction dictionary update information 20 classified by item. It is a figure which shows an example of the threshold value calculation method update information. It is a flowchart which shows the process of a character recognition method.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Character recognition apparatus 10 Document to be recognized 11 Image input part 12 Character recognition part 13 Item classification | category part 14 Item correction | amendment dictionary part 15 Similarity calculation part 16 Date acquisition part 17 Threshold calculation part 18 Correction control part 19 Recognition result 20 Correction according to item Dictionary update information 21 Item-specific correction dictionary update unit 22 Threshold value calculation method update information 23 Threshold value calculation method update unit 30 Character string classification result 31 Keyword dictionary information 32 Rule information 33 Item-specific correction dictionary information 34 Address correction dictionary information 35 Postal code correction dictionary Information 36 URL correction dictionary information

Claims (12)

Storage means for storing correction dictionary information representing a correction dictionary including character strings classified for each preset item in association with update date information representing a date when each item was updated;
Recognizing means for recognizing a character string from image data representing an image including a character string to be recognized;
Classifying means for classifying the character string recognized by the recognition means into any of the preset items;
Among the correction dictionary items indicated by the correction dictionary information stored by the storage unit, for each character string of the same classification item that is the same item as the item into which the classified character string classified by the classification unit is classified, A similarity calculating means for calculating a similarity indicating a degree similar to the classified character string, and determining a character string indicating the maximum value among the calculated similarity values as a similar character string;
Based on a recognition date that is a date when the character string is recognized by the recognition unit and an update date indicated by update date information associated with the same classification item, the character string included in the image is stored in the correction dictionary. A threshold value calculation means for calculating a threshold value indicating a degree of whether or not there is a high possibility of existence;
Replacement means for controlling whether or not to replace the classified character string with the similar character string according to the maximum value of the similarity calculated by the similarity calculation means and the threshold value calculated by the threshold value calculation means And a character recognition device.   The character recognition device according to claim 1, wherein the threshold value calculation means calculates a threshold value according to a period determined by the certification date and the update date according to a calculation formula set in advance for each item. .   The replacement means replaces the classified character string with the similar character string when the maximum value of the similarity is equal to or greater than the threshold value, and replaces the classified character string when the maximum value of the similarity is less than the threshold value. The character recognition apparatus according to claim 1, wherein the character recognition device is not replaced with the similar character string.   The replacement means calculates a difference between the maximum value of the similarity and the threshold, compares the calculated difference value with a predetermined difference value, and if the difference value is equal to or greater than the predetermined difference value, 3. The classified character string is replaced with the similar character string, and when the difference value is less than the predetermined difference value, the classified character string is not replaced with the similar character string. The character recognition device described in 1.   The replacement means calculates a ratio between the maximum value of the similarity and the threshold value, compares the calculated ratio with a predetermined ratio, and if the ratio is equal to or greater than the predetermined ratio, The character recognition apparatus according to claim 1 or 2, wherein the character string is replaced with a similar character string, and the recognized character string is not replaced with the similar character string when the ratio is less than the predetermined ratio.   It further comprises correction dictionary update means for obtaining update information for updating the correction dictionary from the outside and changing the correction dictionary information and update date information stored in the storage means based on the update information. The character recognition device according to any one of claims 1 to 5.   The update information includes item information indicating an item to be updated, position information indicating a position in an item of a character string to be updated, and character string information indicating a character string after update. Item 7. The character recognition device according to Item 6. The threshold value calculation means stores calculation formula information representing a calculation formula for calculating a threshold value for each of the preset items,
It further includes threshold value calculation method updating means for acquiring second update information for updating the calculation formula information from outside and changing the calculation formula information based on the second update information. The character recognition apparatus as described in any one of Claims 1-7.   9. The character recognition device according to claim 8, wherein the second update information includes item information indicating an item to be updated and information indicating a calculation formula after update or a parameter included in the calculation formula. . Computer
Storage means for storing correction dictionary information representing a correction dictionary including character strings classified for each preset item in association with update date information representing a date when each item was updated;
Recognizing means for recognizing a character string from image data representing an image including a character string to be recognized;
Classifying means for classifying the character string recognized by the recognition means into any of the preset items;
Among the correction dictionary items indicated by the correction dictionary information stored by the storage unit, for each character string of the same classification item that is the same item as the item into which the classified character string classified by the classification unit is classified, A similarity calculating means for calculating a similarity indicating a degree similar to the classified character string, and determining a character string indicating the maximum value among the calculated similarity values as a similar character string;
Based on a recognition date that is a date when the character string is recognized by the recognition unit and an update date indicated by update date information associated with the same classification item, the character string included in the image is stored in the correction dictionary. A threshold value calculation means for calculating a threshold value indicating a degree of whether or not there is a high possibility of existence;
Replacement means for controlling whether or not to replace the classified character string with the similar character string according to the maximum value of the similarity calculated by the similarity calculation means and the threshold value calculated by the threshold value calculation means Character recognition program to function as   The computer-readable recording medium which recorded the character recognition program of Claim 10. A recognition step for recognizing a character string from image data representing an image including the character string to be recognized;
A classification step of classifying the character string recognized by the recognition step into any of preset items;
Correction dictionary information stored by storage means for storing correction dictionary information representing a correction dictionary including a character string classified for each preset item in association with update date information representing a date when each item was updated The degree of similarity to the classified character string is indicated for each character string of the same classification item that is the same item as the item into which the classified character string classified by the classification step is classified. A similarity calculation step of calculating a similarity and determining a character string indicating the maximum value among the calculated similarity values as a similar character string;
Based on the recognition date, which is the date when the character string is recognized by the recognition step, and the update date indicated by the update date information associated with the same classification item, the character string included in the image is stored in the correction dictionary. A threshold calculation step for calculating a threshold indicating a degree of whether or not there is a high possibility of existence;
A replacement step for controlling whether to replace the classified character string with the similar character string according to the maximum value of the similarity calculated in the similarity calculation step and the threshold value calculated in the threshold value calculation step. The character recognition method characterized by including these.

Images