JP5387378B2 - Character identification device and character identification method - Google Patents

Description

  The present invention relates to a character identification device and a character identification method.

  For example, in the integration of computer systems due to mergers of cities, towns and villages, it is necessary to process characters that have been processed separately by a plurality of computer systems in a new computer system. In this case, identification work for integrating a plurality of different characters into one character is required in the design stage of a new computer system. The identification operation is an operation in which an operator visually confirms a plurality of characters and determines whether the plurality of characters can be handled as the same character.

  For example, it is necessary to identify characters that are not defined in JIS, in other words, external characters. In addition, identification work for external characters and characters defined in JIS is required. Such external characters are often used for personal names and place names, for example.

  Such identification work is performed, for example, by printing a list of characters that are the object of identification work and searching for characters to be identified while visually checking all the characters. At this time, identification is made efficient by using a type OCR technique for recognizing the printed characters by OCR.

  In the optical character reader, after the character image is divided into n × n blocks, feature vectors of each block are extracted, candidate character groups are searched by comparing the feature blocks with the dictionary, and candidate characters are searched. Is divided into radicals, and when it is determined that the candidate character can be divided into radicals, the character image is divided into a plurality of radical parts, and image parts corresponding to the radicals are divided. It has been proposed to search for a candidate character for each radical portion by processing and to search for a kanji group having the candidate character for each radical at the radical.

  In the character recognition system, when a reject character or a misidentified character in the recognition result is corrected, the feature amount of the reject character or the misidentified character pattern and the feature amount in the recognition dictionary corresponding to the correct character of the reject character or the misidentified character It has been proposed to generate a new feature quantity by combining with and replace the new feature quantity with the feature quantity in the recognition dictionary or add it to the recognition dictionary.

Japanese Patent Laid-Open No. 4-205078 Japanese Patent Laid-Open No. 2-186484

  When visually checking a list of printed characters, if there are thousands of external characters to be identified, the operation is very complicated. Similarly, when there are two or more computer systems to be integrated, it is an extremely complicated and difficult task.

  Even when the type OCR technique is used, candidate characters are extracted by recognizing the entire character, so that the accuracy of the candidate characters is low and only a material that can be referred to to some extent is obtained. In other words, if an appropriate candidate character cannot be obtained, the operator must visually check the list of characters, and the specific gravity is large.

  An object of the present invention is to provide a character identification device that can obtain candidate characters for identification with high accuracy.

  The disclosed character identification device includes a character recognition storage unit, a character shape element storage unit, an external character storage unit, an external character shape element storage unit, a candidate character list generation unit, and an OCR recognition unit. The character recognition storage unit stores a dot pattern of characters. The character element storage unit includes an arrangement pattern indicating the arrangement of radicals for the characters stored in the character recognition storage unit, radical character information including radical character codes indicating radicals, and parts other than radicals The character shape element information of the portion including the partial character code representing the character is stored. The external character storage unit stores a dot pattern of an external character that is a character not included in the standardized character represented by a character code representing a predetermined character. The external character-shaped element storage unit includes an arrangement pattern indicating the arrangement of the radical, a radical character-shaped element information including a radical character code indicating the radical, and a portion other than the radical for the external character stored in the external character storage unit. The character shape element information of the portion including the partial character code representing the character is stored. The OCR recognition unit stores character recognition for an external character to be processed selected from the external character storage unit based on the external character dot pattern stored in the external character storage unit and the character dot pattern stored in the character recognition storage unit. A first candidate character that identifies an external character to be processed is extracted from the character stored in the section. The candidate character list generation unit, for the external character to be processed, the radical character element information for the character stored in the character element storage unit and the radical character element for the external character stored in the external character element storage unit Based on the information, the second candidate character for identifying the external character to be processed is extracted from the character stored in the character element storage unit, and the character stored in the character element storage unit for the external character to be processed The external character to be processed is identified from the characters stored in the glyph element storage unit based on the character element information of the part about and the character element information of the part about the external character stored in the external character element storage unit A third candidate character is extracted.

  According to the disclosed character identification device, even if there are a large number of external characters to be identified, identification candidate characters can be obtained with high accuracy, reducing the burden on the operator performing identification, A period for constructing a new computer system in which a plurality of computer systems are integrated can be shortened.

It is a figure which shows an example of a structure of a character identification device. It is a figure which shows an example of an external character file and an external character form element storage file. It is a figure which shows an example of a character recognition dictionary and a character form element dictionary. It is a figure which shows an example of character identification. It is a figure which shows an example of character identification. It is a figure which shows an example of character identification. It is a figure which shows an example of character identification. It is a figure which shows an example of character identification. It is a figure which shows an example of character identification. It is a figure which shows the processing flow of character identification. It is a figure which shows the processing flow of 1 character identification. It is a figure which shows the processing flow of 1 character identification. It is a figure which shows the processing flow of candidate character list production | generation. It is a figure which shows the processing flow of candidate character learning.

  FIG. 1 is a diagram illustrating an example of the configuration of the character identification device 1.

  The character identification device 1 includes an external character file 2, an external character shape element storage file 3, a character code conversion definition list 4, a display unit 5, and a keyboard 6. The character identification device 1 also includes an identification processing unit 11, a character recognition dictionary 12, an OCR candidate character list 13, a glyph element dictionary 14, a radical candidate character list 15, a partial candidate character list 16, a display candidate character list 17, and an identification. A source / identifier character correspondence list 18 is provided. The identification processing unit 11 includes an OCR recognition unit 111, a candidate character list generation unit 112, a display candidate character list generation unit 113, and a character information learning unit 114.

  In the character identification device 1, the external character file 2 and the external character shape element storage file 3 are external character data sets for storing data of external characters (kanji characters) to be processed. The character to be processed may be a character other than an external character. An external character file 2 and an external character element storage file 3 are prepared in advance. The external character file 2 and the external character shape element storage file 3 store data corresponding to each other, as will be described later.

  The external character file 2 stores an external character dot pattern for each external character. An external character is a character that is not included in a standardized character represented by a character code that represents a predetermined character. The character code is a unique number uniquely assigned to each character or symbol so that the character or symbol is handled by the computer. The character code is, for example, a JIS code. An external character is a character that is not represented by, for example, a JIS code. A dot pattern is data representing a character by a black pattern by giving a white (= 0) or black (= 1) value to each pixel in the character display area.

  The external character shape element storage file 3 stores an arrangement pattern, radical character element information, and character element information of a portion other than the radical for the external characters stored in the external character file 2. The arrangement pattern indicates the arrangement of radicals. The radical character element information of the radical includes a radical character code representing the radical. The character element information of the part other than the radical includes a partial character code representing the part other than the radical.

  Here, the character-shaped element of a character is a radical and a component other than the radical (hereinafter simply referred to as “part”). The character shape is the shape of the entire character, the shape of the radical, or the shape of a component other than the radical. The radical is defined as a reference when classifying kanji, which is one of the glyph elements constituting a character (in other words, kanji). A bias is one of the elements that make up a kanji font and can be broken down into left, right, up, down, inside and outside parts. The radical is a common part of each part of the kanji that serves as a guideline for the kanji arrangement, and includes, for example, a bias, a sword, and a crown. The portion is a component of a character other than the radical, and is a portion obtained by removing the radical of the character from the character. The radical is uniquely specified by the radical character code. The part is uniquely specified by the partial character code.

  The arrangement pattern indicates a position where the radical is arranged. In other words, the radical is assigned an identification number indicating an arrangement pattern as follows, for example, depending on the position where the radical is arranged.

  The arrangement pattern “1” represents “hen”. “Hen” is located on the left side of the kanji divided into right and left. The arrangement pattern “2” represents “production”. “Tsukuri” is located on the right side of the kanji divided into right and left. The arrangement pattern “3” represents “kanmuri”. “Kamuri” is located above the kanji divided into upper and lower. The arrangement pattern “4” represents “ashi”. “Ashi” is located below the kanji divided into upper and lower. The arrangement pattern “5” represents “sag”. The “sag” is positioned so as to sag from the top to the bottom left. The arrangement pattern “6” represents “Nyo”. “Nyo” is located in the shape of “hen” and “ashi” combined from left to bottom. The arrangement pattern “7” represents “Kamae”. “Kamae” is positioned so as to surround the outside.

  In the character identification device 1, the character recognition dictionary 12 and the character form element dictionary 14 are character data sets that store character (Kanji) data for identifying an external character to be processed. The character recognition dictionary 12 and the character form element dictionary 14 are prepared in advance. The character form element dictionary 14 is updated by learning processing, as will be described later. The character recognition dictionary 12 and the character form element dictionary 14 store data corresponding to each other, as will be described later.

  The character recognition dictionary 12 stores a dot pattern of characters. The characters stored in the character recognition dictionary 12 are standardized characters, in other words, characters represented by JIS codes. The characters stored in the character recognition dictionary 12 may include standardized characters and external characters.

  The character form element dictionary 14 stores an arrangement pattern, radical character form element information, and partial character form element information for characters stored in the character recognition dictionary 12. As described above, the arrangement pattern indicates the arrangement of radicals. The radical shape element information of the radical includes the radical character code representing the radical as described above. As described above, the character-shaped element information of the part includes a partial character code representing a part other than the radical.

  The OCR recognition unit 111 reads an external character from the external character file 2 and sets it as a processing target. The OCR recognition unit 111 extracts a first candidate character that identifies the external character that is the processing target for the external character that is the processing target selected from the external character file 2. The first candidate character is extracted from the character stored in the character recognition dictionary 12 based on the dot pattern of the external character stored in the external character file 2 and the dot pattern of the character stored in the character recognition dictionary 12.

  Specifically, the OCR recognition unit 111 performs OCR processing on the dot pattern of the external character to be processed and the dot pattern of the character stored in the character recognition dictionary 12. As a result, the OCR recognition unit 111 extracts, from the character dot patterns stored in the character recognition dictionary 12, a character that is similar to the external character dot pattern to be processed as a whole.

  The OCR recognition unit 111 sets the extracted dot pattern character as a first candidate character for identifying an external character to be processed. One or a plurality of first candidate characters are extracted. The OCR recognition unit 111 stores the first candidate character in the OCR candidate character list 13. Thereby, the OCR candidate character list 13 is generated for the external characters to be processed. The OCR recognition unit 111 notifies the display candidate character list generation unit 113 of the generation of the OCR candidate character list 13.

  The OCR recognition unit 111 notifies the candidate character list generation unit 112 of the external characters to be processed. In response to this, the candidate character list generation unit 112 refers to the external character shape element storage file 3 and sets the radical character shape of the external character stored in the external character shape element storage file 3 for the external character to be processed. Read element information and character information of part.

  The candidate character list generation unit 112 extracts a second candidate character that identifies the external character that is the processing target for the external character that is the processing target. The second candidate character is based on the radical character element information for the character stored in the character element dictionary 14 and the radical character element information for the external character stored in the external character element storage file 3. Extracted from the characters stored in the glyph element dictionary 14.

  Specifically, the candidate character list generation unit 112 converts the radical character code in the radical character element information for the external character to be processed into the radical character element information for the character stored in the character element dictionary 14. Compare with the radical code in. Thereby, the candidate character list generation unit 112 extracts a character having the same radical character code as the radical character code of the external character to be processed from the characters stored in the character form element dictionary 14.

  Candidate character list generation unit 112 sets the extracted character as a second candidate character for identifying an external character to be processed. One or a plurality of second candidate characters are extracted. The candidate character list generation unit 112 stores the second candidate character in the radical candidate character list 15. Thereby, the radical candidate character list 15 is generated for the external characters to be processed. The candidate character list generation unit 112 notifies the generation of the radical candidate character list 15 to the display candidate character list generation unit 113.

  In addition, the candidate character list generation unit 112 extracts a third candidate character that identifies the external character that is the processing target for the external character that is the processing target. The third candidate character is a character element based on the character element information of the part stored in the character element dictionary 14 and the character element information of the part of the external character stored in the external character element storage file 3. Extracted from characters stored in the dictionary 14.

  Specifically, the candidate character list generation unit 112 converts the partial character code in the character element information of the part for the external character to be processed into the partial character in the character element information of the part for the character stored in the character element dictionary 14. Compare with code. Thereby, the candidate character list generation unit 112 extracts a character having the same partial character code as the partial character code of the external character to be processed from the characters stored in the character form element dictionary 14.

  The candidate character list generation unit 112 sets the extracted character as a third candidate character for identifying an external character to be processed. One or a plurality of third candidate characters are extracted. The candidate character list generation unit 112 stores the third candidate character in the partial candidate character list 16. Thereby, the radical candidate character list 16 is generated for the external characters to be processed. The candidate character list generation unit 112 notifies the generation of the radical candidate character list 16 to the display candidate character list generation unit 113.

  When the generation of the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16 is notified, the display candidate character list generation unit 113 refers to these. Based on the reference result, the display candidate character list generation unit 113 notifies the character information learning unit 114 of a character that is common to each of the first to third candidate characters. To do.

  When there is no character common to each of the first candidate character to the third candidate character, the display candidate character list generation unit 113 displays the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate characters. Based on the list 16, a display candidate character list 17 is generated. The display candidate character list 17 is generated by merging the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16. The display candidate character list generation unit 113 displays the display candidate character list 17 on the display unit 5. The display unit 5 displays the first candidate character to the third candidate character. Instead of this, the display unit 5 may be an output unit capable of outputting the display candidate character list 17.

  Here, the display candidate character list generation unit 113 assigns priorities to the first candidate character to the third candidate character according to the degree to which the candidate characters are included redundantly. For example, a character that is duplicated in any two of the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16 has a higher priority than the other characters. The other characters are characters that exist only in any one of the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16. The difference in the priority order is represented by, for example, a display color difference, a display character thickness, or the like.

  Thereby, the operator who has seen the display candidate character list 17 displayed on the display unit 5 can easily select the character for identifying the character to be processed. For example, in response to an operator input, the keyboard 6 serving as an input device inputs a selection input for designating a candidate character displayed on the display unit 5 to the character information learning unit 114.

  Note that the display candidate character list 17 may be generated and displayed even when there is a character common to each of the first to third candidate characters. In this case, the priority order of the common characters is the highest.

  The character information learning unit 114 determines an identification target character that is a character for identifying an external character to be processed from the first candidate character to the third candidate character. Therefore, the external character to be processed is the same as another character, in other words, an identification source character identified by another character. The first to third candidate characters are characters that may be identified as other characters, in other words, characters that may be the identification target characters.

  Specifically, as described above, the character information learning unit 114 determines that characters common to each of the first candidate character to the third candidate character are received in response to the notification from the display candidate character list generation unit 113. If it exists, the common character is determined as the identification target character. Further, the character information learning unit 114 determines an identification destination character based on a selection input for designating a candidate character displayed on the display unit 5 input from the keyboard 6. This is a case where there is no character common to each of the first to third candidate characters.

  The character information learning unit 114 uses an external character to be processed as an identification source character, generates a character correspondence list 18 that associates the identification source character and the identification destination character, and outputs this as the character code conversion definition list 4. The character correspondence list 18 or the character code conversion definition list 4 is, for example, a list that associates an external character or external character code that is an identification source character with a character or JIS code that is an identification destination character.

  Further, the character information learning unit 114 identifies the external character to be processed as the identification source character, identifies the radical shape element information of the identification source character or the character shape element information of the portion as the learning element information about the identification destination character. It is added to the radical shape element information or the partial shape element information of the first character. As a result, the candidate character list generation unit 112 extracts the second candidate character and the third candidate character based on the added radical learning element information and the partial learning element information. Thereby, when the identification process is performed again for the identification source character once identified as the identification destination character, the identification source character once identified as the identification destination character is actually recognized as a candidate character.

  FIG. 2 is a diagram illustrating an example of the external character file 2 and the external character element storage file 3.

  The external character file 2 includes character data 21 and a dot pattern 22 corresponding to the character data 21. The character data 21 and the dot pattern 22 are provided for each of the external characters included in the external character file 2.

  The character data 21 includes an identification source character and a storage destination address of the identification source character. The identification source character is a character identified as a certain character, for example, an external character. The identification source character may be represented by, for example, identification information (external character code) that uniquely defines the external character that is the identification source character. The storage destination address is an address where the dot pattern 22 of the identification source character is stored. The dot pattern 22 is a pattern in which the identification source character is represented by a set of dots.

  The character “Mine” is originally a character that is standardized by the JIS code and is not an external character, but in this specification, for the purpose of explanation, it is used as an example of a standardized character and an example of an external character. And In other words, “Mine” is used as the standardized character “Mine” or the external character “Mine”.

  The external character form element storage file 3 includes a plurality of external character form element information 31A to 31D. The external character element information 31 </ b> A to 31 </ b> D is provided for each external character included in the external character file 2. The external character element information 31A to 31D includes an identification source character, an arrangement pattern indicating an arrangement of a radical of the identification source character, a radical character element information including a radical character code indicating a radical of the identification source character, And character-shaped element information of a part including a partial character code representing a part of the identification source character. The identification source character may be represented by, for example, identification information (external character code) that uniquely defines the external character that is the identification source character. The external character file 2 and the external character element storage file 3 are associated by including the same identification source character. In FIG. 2, the arrangement pattern is represented as “arrangement”, the radical character code is represented as “radical”, and the partial character code is represented as “part”.

  For example, the external character shape element information 31A stores the arrangement pattern “1”, the radical character code [gold], and the partial character code [wide] for the identification source character “mine”. In this specification, for example, the radical character code of the radical “gold” is represented as [gold], and the partial character code of the portion “wide” is represented as [wide].

  FIG. 3 is a diagram showing an example of the character recognition dictionary 12 and the character form element dictionary 14.

  The character recognition dictionary 12 includes character data 121 and a dot pattern 122 corresponding to the character data 121. The character data 121 and the dot pattern 122 are provided for each character included in the character recognition dictionary 12. Characters included in the character recognition dictionary 12 are characters (identification destination characters) for which other characters are desired to be identified. Accordingly, as will be described later, they are candidate characters for identification destination characters. The character included in the character recognition dictionary 12 is a character represented by a character code representing a predetermined character, such as a JIS code, in other words, a standardized character.

  The characters included in the character recognition dictionary 12 may be characters that are not standardized characters, in other words, external characters. Therefore, the characters included in the character recognition dictionary 12 include at least standardized characters, and may include external characters in addition to these.

  The character data 121 includes an identification destination character and a storage destination address of the identification destination character. The identification target character is a character with which another character is identified, for example, a standardized character. For example, the identification destination character may be represented by identification information (character code) that uniquely defines the identification destination character. The storage destination address is an address where the dot pattern 122 of the identification destination character is stored. The dot pattern 122 is a pattern in which the identification target character is represented by a set of dots.

  The glyph element dictionary 14 includes a plurality of glyph element structures 141. The character element structure 141 is provided for each character included in the character recognition dictionary 12. The glyph element structure 141 is a radical character element information including an identification target character, the number of learning characters, an arrangement pattern indicating the arrangement of the radical of the identification target character, and a radical character code representing the radical of the identification target character. And character shape element information of a part including a partial character code representing the part of the identification target character. For example, the identification destination character may be represented by identification information (character code) that uniquely defines the identification destination character. The character recognition dictionary 12 and the character form element dictionary 14 are associated by including the same identification target character. In FIG. 3, the arrangement pattern is represented as “arrangement”, the radical character code is represented as “radical”, and the partial character code is represented as “part”.

  In practice, the glyph element structure 141 includes radical rows and partial rows, as shown in FIG. The radical row includes the number of learned characters, the arrangement pattern, and a plurality of radical character codes for the radical. The learning character number for the radical is the number of radical character codes included in the radical row. The row of the portion includes the number of learned characters, the arrangement pattern, and a plurality of partial character codes for the portion. The number of learning characters for a portion is the number of partial character codes included in the portion line. All the arrangement patterns have the same value in the same character element structure 141.

  A plurality of radical character codes and arrangement patterns are stored in order from the top in an array [0], an array [1],. An array in which a plurality of radical codes and arrangement patterns are not stored is “NULL (empty)”. The partial character code and the arrangement pattern are stored in order from the top in the array [0], the array [1],. An array in which the partial character code and the arrangement pattern are not stored is “NULL (empty)”.

  For example, the character element structure 141 stores the learning character number “1” for the radical “gold” of the identification target character “mine”, and the arrangement pattern “1” and the radical character code [gold] in the array [0]. Is stored. The character element structure 141 stores the learning character number “1” for the part “wide” of the identification target character “mine”, and the arrangement pattern “1” and the partial character code [wide] in the array [0]. To do.

  Here, the arrangement pattern and radical character code stored in the array [0] are the original (default) radical arrangement pattern and radical character code of the identification target character stored in the character element structure 141. Represent. The arrangement pattern and partial character code stored in the array [0] represent the original (default) partial arrangement pattern and partial character code of the identification target character stored in the character element structure 141. In other words, the arrangement pattern, radical character code, and partial character code stored in the array [0] are the default values of the identification target character stored in the character element structure 141, and are stored in advance.

  On the other hand, the arrangement patterns and radical character codes stored after the array [1] are acquired by a learning process based on the character identification process. Further, the arrangement patterns and partial character codes stored after the array [1] are acquired by a learning process based on the character identification process. In other words, the arrangement pattern, radical character code, and partial character code stored after the array [1] are values representing the identification source character identified by the identification destination character stored in the character element structure 141. , Additionally stored as a result of the learning process.

  FIG. 4 is a diagram illustrating an example of character identification.

  In the example illustrated in FIG. 4, as described above, the character “Mine” is a character standardized by the JIS code and is also an external character. In this case, it is assumed that the external character “Mine”, which is the first identification character, is a variant of the character “Mine” standardized by the JIS code. An alloface means a font that has the same spelling but a different font (dot pattern). In addition, the second to fourth identification source characters are not characters standardized by the JIS code but external characters. The following FIGS. 5 to 9 will be described using the example of FIG.

  For example, the first identification source character (external character “mine”) is an external character composed of a radical “gold” and a portion “wide”. In this case, since the “dot pattern” is similar, the character “Mine” is extracted as the first candidate character and included in the OCR candidate character list 13. The similarity of “dot pattern” will be described later. In addition, since the “radical” matches, the character “mine” is extracted as the second candidate character and included in the radical candidate character list 15. Also, since “part” matches, the character “mine” is extracted as the third candidate character and included in the partial candidate character list 16. As a result, since the character “Mine” is included in common among the first to third candidate characters, the first identification source character is identified as the identification destination character “Mine”.

  The second identification source character is an external character composed of a radical “gold” and a portion “廣”. In this case, since the “dot pattern” is not similar, the character “mine” is not extracted as the first candidate character and is not included in the OCR candidate character list 13. In addition, since the “radical” matches, the character “mine” is extracted as the second candidate character and included in the radical candidate character list 15. On the other hand, since the “part” does not match, the character “mine” is not extracted as the third candidate character and is not included in the partial candidate character list 16. As a result, since the character “mine” is included in the second candidate character, the second identification source character is identified as the identification destination character “mine” in accordance with the selection input to the character information learning unit 114.

  The third identification source character is an external character composed of a radical “stone” and a portion “wide”. In this case, since the “dot pattern” is not similar, the character “mine” is not extracted as the first candidate character and is not included in the OCR candidate character list 13. In addition, since the “radical” does not match, the character “mine” is not extracted as the second candidate character and is not included in the radical candidate character list 15. However, since the “part” matches, the character “mine” is extracted as the third candidate character and included in the partial candidate character list 16. As a result, since the character “Mine” is included in the third candidate character, the third identification source character is identified as the identification destination character “Mine” according to the selection input to the character information learning unit 114.

  The fourth identification source character is an external character composed of a radical “stone” and a portion “廣”. In this case, since the “dot pattern” is not similar, the character “mine” is not extracted as the first candidate character and is not included in the OCR candidate character list 13. In addition, since the “radical” does not match, the character “mine” is not extracted as the second candidate character and is not included in the radical candidate character list 15. In addition, since the “part” does not match, the character “mine” is not extracted as the third candidate character and is not included in the partial candidate character list 16.

  In this way, the fourth identification source character is initially, in other words, the character “Mine” is not extracted as the first to third candidate characters before the learning process. Does not match. However, as will be described later with reference to FIG. 8, as a result of the learning process, the character “mine” is included in the second and third candidate characters, and therefore the fourth identification source character is sent to the character information learning unit 114. According to the selection input, the identification target character “Mine” is identified.

  5 to 9 are diagrams illustrating an example of character identification. In particular, FIGS. 5 to 8 show the case where the first identification source character to the fourth identification source character are identified in this order. FIG. 9 shows a case where the first identification source character to the fourth identification source character are identified and then the fourth identification source character is identified again.

  FIG. 5 shows an identification process for a first identification source character (external character “Mine”) composed of a radical “Gold” and a portion “Wide”.

  As described above, the character “Mine” is a character standardized by the JIS code, and is stored in the character element structure 141 of the character element dictionary 14 as shown in FIG. On the other hand, the external character “Mine” which is a variant of the character “Mine” is not identified in any character. Therefore, the external character “Mine”, which is a variant of the character “Mine”, is stored in the external character element information 31A of the external character element storage file 3 as shown in FIG.

  The external character file 2 stores a dot pattern for the first identification source character that is an external character. The external character form element storage file 3 stores external character form element information 31 </ b> A for the first identification source character stored in the external character file 2. In this case, the external character element information 31A includes the arrangement pattern “1”, the radical character information including the radical character code indicating the radical “gold”, and the part “wide” for the first identification source character. The character shape element information of the portion including the partial character code representing the character is stored.

  For example, the OCR recognition unit 111 performs a character recognition process using the character recognition dictionary 12 for the first identification source character composed of the radical “gold” and the portion “wide”. Thereby, the OCR recognition unit 111 extracts a plurality of characters having similar “dot patterns” as the first candidate characters for the first identification source character, and stores them in the OCR candidate character list 13. The character extracted by the similarity of “dot pattern” includes the character “Mine”.

  The similarity of “dot pattern” includes, for example, the case where the values of pixels equal to or higher than a predetermined ratio match in two dot patterns to be compared. The ratio can be determined empirically and is a relatively small value. Further, the similarity of the “dot pattern” may include a case where the radicals are matched or similar, a case where the parts are matched or similar, and the like. Thereby, as shown in FIGS. 5 to 9, a relatively large number of characters can be extracted as similar characters. The dot patterns of the characters “Mine” and the external character “Mine” standardized by the JIS code are similar to each other because they are different characters.

  In addition, the candidate character list generation unit 112 compares the character elements using the character element dictionary 14 for the first identification source character, in particular, for the radical “gold” of the first identification source character. As a result, the candidate character list generation unit 112 extracts a plurality of characters having the same “radical” as the second candidate character for the first identification source character, and stores them in the radical candidate character list 15. . The character extracted by the match of “radical” includes the character “Mine”.

  At this time, in the character element structure 141 which is the learning data of the identification target character “Mine” in the character element dictionary 14, the radical “gold” of the first identification character is stored in the array [0]. Therefore, in this case, the candidate character list generation unit 112 stores the character “Mine” in the “candidate character” field instead of the “learning candidate” field in the radical candidate character list 15.

  In addition, the candidate character list generation unit 112 compares the character elements using the character element dictionary 14 for the first identification source character, in particular, for the portion “wide” of the first identification source character. As a result, the candidate character list generation unit 112 extracts a plurality of characters that match “part” as the third candidate character for the first identification source character, and stores them in the partial candidate character list 16. The character extracted by the match of “part” includes the character “Mine” (process # 51).

  At this time, in the character element structure 141 which is the learning data of the identification target character “Mine” in the character element dictionary 14, the portion “Wide” of the first identification source character is stored in the array [0]. Therefore, in this case, the candidate character list generation unit 112 stores the character “Mine” in the “candidate character” field instead of the “learning candidate” field in the radical candidate character list 15.

  Thereafter, the display candidate character list generation unit 113 includes the same character “Mine” in each of the OCR candidate character list 13, radical candidate character list 15, and partial candidate character list 16. The character “Mine” is determined (Process # 52). As a result, the identification character “Mine” for the identification source character “Mine” which is an external character is determined.

  Therefore, the character information learning unit 114 determines the first identification source character when the same character “Mine” is included in each of the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16. The learning process for learning as the learning data of the identified character “Mine” is not performed (process # 53). In other words, the display candidate character list 17 is not generated and is not displayed on the display unit 5.

  Specifically, in this case, the radical code [gold] of the identification source character “Mine” is the same as the radical code [gold] stored in the array [0] of the character element structure 141. Therefore, the radical code [gold] of the identification source character “Mine” is not added to the character-shaped element structure 141 as learning data of the identification destination character “Mine”. Further, the partial character code [Wide] of the identification source character “Mine” is the same as the radical character code [Wide] stored in the array [0] of the character element structure 141. Therefore, the partial character code [Wide] of the identification source character “Mine” is not added to the character-shaped element structure 141 as learning data of the identification destination character “Mine”.

  From the above, in this case, the first identification source character is not added to the glyph element structure 141 as learning data of the identification target character “Mine”. Therefore, the glyph element structure 141 of the identification target character “Mine” has the arrangement pattern and the radical “Gold” of the identification target character “Mine” in the array [0], as before the identification of the first identification source character. ”And a part“ Wide ”of the identification character“ Mine ”.

  FIG. 6 shows an identification process for the second identification source character composed of the radical “gold” and the part “廣”. As described above, the second identification source character is not a character standardized by the JIS code but an external character.

  In this case, the external character element information 31B includes the arrangement pattern “1”, the radical character information including the radical character code indicating the radical “gold”, and the part “廣” for the second identification source character. The character shape element information of the portion including the partial character code representing the character is stored.

  For example, the OCR recognizing unit 111 performs a character recognition process on the second identification source character using the character recognition dictionary 12 to obtain “dot pattern” as the first candidate character for the second identification source character. Are extracted and stored in the OCR candidate character list 13. In addition, the candidate character list generation unit 112 compares the glyph elements for the radical “gold” of the second identification source character by using the glyph element dictionary 14, so that the second identification source character for the second identification source character is compared. As the second candidate character, a plurality of characters having the same “radical” are extracted and stored in the radical candidate character list 15. In addition, the candidate character list generation unit 112 compares the glyph elements with respect to the part “部分” of the second identification source character by using the glyph element dictionary 14, thereby performing the third identification of the second identification source character. As the candidate characters, a plurality of characters with the same “part” are extracted and stored in the partial candidate character list 16 (process # 61).

  At this time, in the character element structure 141 that is the learning data of the identification target character “Mine” in the character element dictionary 14, the radical “gold” of the second identification source character is stored in the array [0]. Therefore, in this case, the candidate character list generation unit 112 stores the character “Mine” in the “candidate character” field instead of the “learning candidate” field in the radical candidate character list 15.

  Thereafter, the display candidate character list generation unit 113 determines the identification target character because the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16 do not include the same character in common. Process. Specifically, the display candidate character list generating unit 113 generates a display candidate character list 17 based on the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16, and the display unit 5 To display.

  The operator who sees this inputs, for example, the character “Mine” from the keyboard 6 as an instruction to select the identification target character. Note that the character “mine” may be input by selecting from the characters displayed in the portion corresponding to the radical candidate character list 15 in the display candidate character list 17. The same applies to FIGS. 8 to 9. In response to this instruction, the character information learning unit 114 determines the character “mine” as the identification target character (processing # 62). As a result, the identification target character “Mine” for the second identification source character is determined.

  Thereafter, the character information learning unit 114 performs a learning process of learning the second identification source character as learning data of the determined identification destination character “Mine” (processing # 63). As a result, the character information learning unit 114 adds the learning element array to the character element structure 141 of the determined identification target character “Mine” stored in the character element dictionary 14.

  In this case, the radical character code [gold] of the second identification source character is the same as the radical character code [gold] stored in the array [0] of the character element structure 141. Therefore, the radical character code [gold] of the second identification source character is not added to the character element structure 141 as learning data of the identification target character “Mine”. On the other hand, the partial character code [廣] of the second identification source character is different from the radical character code [wide] stored in the array [0] of the character element structure 141. Therefore, the partial character code [廣] of the second identification source character is added to the character-shaped element structure 141 as learning data of the identification destination character “Mine”.

  From the above, in this case, the character-shaped element structure 141 of the identification target character “Mine” has the arrangement pattern “1” in the storage field of “part” of the array [1] by learning the second identification source character. The part “部分” of the second identification source character is stored. In addition, since one partial character code or the like is newly stored in the “part” storage field of the array [1], the number of learning characters for “part” is set to “2”. At this time, the storage field of “radical” in the array [1] is set to “NULL (empty)” because there is no radical character code to be stored. In addition, since a new radical character code or the like is not stored in the “radical” storage field of the array [1], the number of learning characters for “radical” is also kept at “1”.

  As a result of this learning process, the second identification source character is recognized as a candidate character of the character “Mine” by learning the character element “廣” as the character element of the identification target character “Mine”. . Thus, the character having the character form element “要素” is set as a candidate character for the character “Mine”.

  FIG. 7 shows an identification process for the third identification source character composed of the radical “stone” and the portion “wide”. As described above, the third identification source character is not a character standardized by the JIS code but an external character.

  In this case, the external character element information 31C includes the arrangement pattern “1”, the radical character element information including the radical character code indicating the radical “stone”, and the part “wide” for the third identification source character. The character shape element information of the portion including the partial character code representing the character is stored.

  For example, the OCR recognition unit 111 performs a character recognition process on the third identification source character using the character recognition dictionary 12, so that “dot pattern” is set as the first candidate character for the third identification source character. Are extracted and stored in the OCR candidate character list 13. In addition, the candidate character list generation unit 112 compares the glyph elements for the radical “stone” of the third identification source character by using the glyph element dictionary 14, so that the third identification source character for the third identification source character is compared. As the second candidate character, a plurality of characters having the same “radical” are extracted and stored in the radical candidate character list 15. In addition, the candidate character list generation unit 112 compares the character element with the character element dictionary 14 using the character element dictionary 14 for the part “wide” of the third identifier character, so that a third character character about the third identifier character is displayed. As the candidate characters, a plurality of characters having the same “part” are extracted and stored in the partial candidate character list 16 (process # 71).

  At this time, in the character element structure 141 which is the learning data of the identification target character “Mine” in the character element dictionary 14, the portion “Wide” of the third identification source character is stored in the array [0]. Therefore, in this case, the candidate character list generation unit 112 stores the character “Mine” in the “candidate character” field instead of the “learning candidate” field in the partial candidate character list 16.

  Thereafter, the display candidate character list generation unit 113 determines the identification target character because the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16 do not include the same character in common. Process. Specifically, the display candidate character list generating unit 113 generates a display candidate character list 17 based on the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16, and the display unit 5 To display.

  The operator who sees this inputs, for example, the character “Mine” from the keyboard 6 as an instruction to select the identification target character. In response to this instruction, the character information learning unit 114 determines the character “Mine” as the identification target character (processing # 72). As a result, the identification target character “Mine” for the third identification source character is determined.

  Thereafter, the character information learning unit 114 performs a learning process of learning the third identification source character as learning data of the determined identification destination character “Mine” (processing # 73). As a result, the character information learning unit 114 adds the learning element array to the character element structure 141 of the determined identification target character “Mine” stored in the character element dictionary 14.

  In this case, the radical character code [stone] of the third identification source character is different from the radical character code [gold] stored in the array [0] of the character element structure 141. Therefore, the radical code [stone] of the third identification source character is added to the character-shaped element structure 141 as learning data of the identification destination character “Mine”. On the other hand, the partial character code [wide] of the third identification source character is the same as the radical character code [wide] stored in the array [0] of the character element structure 141. Therefore, the partial character code [Wide] of the third identification source character is not added to the character-shaped element structure 141 as learning data of the identification target character “Mine”.

  From the above, in this case, the character-shaped element structure 141 of the identification target character “Mine” has the arrangement pattern “1” in the storage field of “radical” of the array [1] by learning the third identification source character. The radical “stone” of the third identification source character is stored. In addition, since one radical character code or the like is newly stored in the “radical” storage field of the array [1], the number of learning characters for “radical” is set to “2”. In addition, since a new partial character code or the like is not stored in the “part” storage field of array [1], the number of learning characters for “part” is also kept at “2”.

  As a result of this learning process, the third identification source character is recognized as a candidate character of the character “Mine” by learning the character shape element “Stone” as the character shape element of the identification destination character “Mine”. . Thereby, the character having the character element “stone” is set as a candidate character for the character “Mine”.

  FIG. 8 shows an identification process for the fourth identification source character composed of the radical “stone” and the portion “廣”. As described above, the fourth identification source character is not a character standardized by the JIS code but an external character.

  In this case, the external character element information 31D includes the arrangement pattern “1”, the radical character element information including the radical character code indicating the radical “stone”, and the part “廣” for the fourth identification source character. The character shape element information of the portion including the partial character code representing the character is stored.

  For example, the OCR recognizing unit 111 performs a character recognition process on the fourth identification source character using the character recognition dictionary 12, thereby obtaining “dot pattern” as the first candidate character for the fourth identification source character. Are extracted and stored in the OCR candidate character list 13. In addition, the candidate character list generation unit 112 compares the glyph elements for the radical “stone” of the fourth identification source character by using the glyph element dictionary 14, so that the fourth character for the fourth identification source character is compared. As the second candidate character, a plurality of characters having the same “radical” are extracted and stored in the radical candidate character list 15. In addition, the candidate character list generation unit 112 compares the glyph elements of the fourth identification source character portion “廣” by using the glyph element dictionary 14, thereby performing the third identification of the fourth identification source character. Are extracted as a candidate character, and stored in the partial candidate character list 16 (process # 81).

  At this time, the radical “stone” of the fourth identification source character is stored in the array [1] in the character shape element structure 141 which is the learning data of the identification target character “Mine” in the character shape element dictionary 14. Therefore, in this case, the candidate character list generation unit 112 stores the character “Mine” in the “learning candidate” field, not the “candidate character” field, in the radical candidate character list 15.

  Further, in the character shape element structure 141 which is the learning data of the identification target character “Mine” in the character shape element dictionary 14, the portion “廣” of the fourth identification source character is stored in the array [1]. In this case, the candidate character list generation unit 112 stores the character “Mine” in the “candidate character” field instead of the “candidate character” field in the partial candidate character list 16.

  Thereafter, the display candidate character list generation unit 113 determines the identification target character because the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16 do not include the same character in common. Process. Specifically, the display candidate character list generating unit 113 generates a display candidate character list 17 based on the OCR candidate character list 13, the radical candidate character list 15, and the partial candidate character list 16, and the display unit 5 To display.

  At this time, the character “Mine” is included in both the radical candidate character list 15 and the partial candidate character list 16 in common. Therefore, since the character “Mine” is a candidate character with a high priority, for example, it is displayed in a color different from that of a candidate character with a low priority.

  The operator who sees this inputs, for example, the character “Mine” from the keyboard 6 as an instruction to select the identification target character. In response to this instruction, the character information learning unit 114 determines the character “Mine” as the identification target character (processing # 82). As a result, the identification target character “Mine” for the fourth identification source character is determined.

  Thereafter, the character information learning unit 114 performs a learning process of learning the fourth identification source character as learning data of the determined identification destination character “Mine” (processing # 83). As a result, the character information learning unit 114 adds the learning element array to the character element structure 141 of the determined identification target character “Mine” stored in the character element dictionary 14.

  In this case, the radical character code [stone] of the fourth identification source character is the same as the radical character code [stone] stored in the array [1] of the character element structure 141. Therefore, the radical code [stone] of the fourth identification source character is not added to the character-shaped element structure 141 as learning data of the identification target character “Mine”. Further, the partial character code [の] of the fourth identification source character is the same as the radical character code [廣] stored in the array [1] of the character element structure 141. Therefore, the partial character code [廣] of the fourth identification source character is not added to the character shape element structure 141 as learning data of the identification destination character “Mine”.

  From the above, in this case, the fourth identification source character learning process is executed, but a new radical character code and partial character are stored in any storage field of the character element structure 141 of the identification destination character “Mine”. Since the code or the like is not stored, the number of learning characters for “radical” and “part” is also kept at “2”.

  As a result of this learning process, the fourth identification source character is recognized as a candidate character of the character “Mine” by learning the character element “Stone” as the character element of the identification target character “Mine”. .

  FIG. 9 shows an identification process for a new identification source character composed of the radical “stone” and the portion “廣”. The new identification source character is not a character standardized by the JIS code but an external character. The new identification source character is assumed to be a variant of the second identification source character composed of the above-described radical “gold” and the portion “廣”.

  In this case, the external character-shaped element information 31E includes the arrangement pattern “1”, the radical-shaped element information including the radical character code representing the radical “gold”, and the part “廣” for the new identification source character. The character shape element information of the part including the partial character code to be expressed is stored.

  For example, the OCR recognition unit 111 performs a character recognition process on the new identification source character using the character recognition dictionary 12, so that “dot pattern” is set as the first candidate character for the new identification source character. A plurality of similar characters are extracted and stored in the OCR candidate character list 13. In addition, the candidate character list generation unit 112 compares the glyph elements of the radical “gold” of the new identification source character by using the glyph element dictionary 14, thereby performing the second identification of the new identification source character. A plurality of characters with the same “radical” are extracted as candidate characters and stored in the radical candidate character list 15. Further, the candidate character list generation unit 112 compares the glyph elements with respect to the new identification source character portion “廣” using the glyph element dictionary 14, thereby providing a third candidate for the new identification source character. A plurality of characters that match “part” are extracted as characters and stored in the partial candidate character list 16 (process # 91).

  At this time, the radical “gold” of the fourth identification source character is stored in the array [0] in the character element structure 141 which is the learning data of the identification target character “Mine” in the character element dictionary 14. Therefore, in this case, the candidate character list generation unit 112 stores the character “Mine” in the “candidate character” field instead of the “learning candidate” field in the radical candidate character list 15.

  Further, in the character shape element structure 141 which is the learning data of the identification target character “Mine” in the character shape element dictionary 14, the portion “廣” of the fourth identification source character is stored in the array [1]. In this case, the candidate character list generation unit 112 stores the character “Mine” in the “candidate character” field instead of the “candidate character” field in the partial candidate character list 16.

  Thereafter, the display candidate character list generation unit 113 includes the same character “Mine” in each of the OCR candidate character list 13, radical candidate character list 15, and partial candidate character list 16. Then, the character “mine” is determined (process # 92). As a result, the identification character “Mine” for the identification source character “Mine” which is an external character is determined.

  Therefore, the character information learning unit 114 does not perform a learning process of learning a new identification source character as learning data of the identified identification target character “Mine” (processing # 93), and the display candidate character list 17 is generated. Is not displayed on the display unit 5.

  Specifically, in this case, the radical code [gold] of the identification source character is the same as the radical code [gold] stored in the array [0] of the character element structure 141. Therefore, the radical code [gold] of the identification source character “Mine” is not added to the character-shaped element structure 141 as learning data of the identification destination character “Mine”. Further, the partial character code [廣] of the identification source character is the same as the radical character code [廣] stored in the array [1] of the character element structure 141. Therefore, the partial character code [廣] of the identification source character is not added to the glyph element structure 141 as learning data of the identification destination character “Mine”. From the above, in this case, the new identification source character is not added to the character element structure 141 as learning data of the identification destination character “Mine”.

  FIG. 10 is a flow of character identification processing.

  For example, the operator generates a character element for characters stored in the external character file 2 (step S1). Thereby, the external character shape element storage file 3 corresponding to the external character file 2 is obtained. Thereafter, the identification processing unit 11 checks whether or not the identification process has been completed for all characters (external characters) stored in the external character file 2 (step S2).

  If the identification process has not been completed for all characters (external characters) (No in step S2), the identification processing unit 11 selects and reads out a character pattern for one character from the external character file 2 as a character to be processed, For the character to be processed, the radical character element information and the partial character element information of the radical are read from the external character element storage file 3 (step S3).

  The identification processing unit 11 performs an identification process on the character to be processed (step S4), and executes step S2.

  When the identification process is completed for all characters (external characters) in step S2 (Yes in step S2), the identification processing unit 11 determines whether the characters in the identification source external character file 2 and the identification destination characters correspond to each other. The code conversion definition list 4 is generated (step S5).

  11 and 12 are identification processing flows for one character.

  The OCR recognition unit 111 performs character recognition processing on the character to be processed using the character recognition dictionary 12 (step S11), and generates an OCR candidate character list 13 as a result of character recognition (step S12).

  The candidate character list generation unit 112 checks whether or not there is a radical (step S13). If there is a radical (No in step S13), the candidate character list generation unit 112 generates the radical candidate character list 15 and the partial candidate character list 16 using the glyph elements (step S14). If there is no radical (step S13 Yes), step S14 is omitted.

  Thereafter, the display candidate character list generation unit 113 generates a display candidate character list (step S15). At this time, the character existing in any candidate character list is given high priority as a highly accurate candidate.

  Thereafter, the display candidate character list generation unit 113 checks whether or not there is a character in any candidate character list (step S16). If there is no character that exists in common in any candidate character list (No in step S16), the display candidate character list generation unit 113 displays the candidate character list on the display unit 5 (step S17). Based on the selection input by the operator who sees this, the character information learning unit 114 finalizes the identification destination character (step S18), and causes the character shape element dictionary 14 of the identification destination character to learn character shape element information of the identification source character ( Step S19).

  In step S16, when there is a character that exists in common in any candidate character list (step S16 Yes), steps S17 to S19 are omitted, and the character that exists in common is determined as the identification target character.

  FIG. 13 is a processing flow for generating a candidate character list.

  The candidate character list generation unit 112 checks whether or not all characters in the set of identification target characters have been processed (step S21). When not all the characters in the set of identification target characters have been processed (No in step S21), the candidate character list generation unit 112 uses the glyph element dictionary 14 to match the arrangement pattern and the radical character code. The radical is added to the radical candidate character list 15 (step S22), and when the arrangement pattern and the partial character code match, the portion is added to the partial candidate character list 16 (step S23), and then step S21. Execute.

  In step S21, when all the characters in the set of identification target characters have been processed (step S21 Yes), the process ends.

  FIG. 14 is a processing flow of candidate character learning.

  The character information learning unit 114 adds the arrangement pattern and the radical character code of the identification source character to the radical learning list of the identification target character in the character shape dictionary 14, and increments the number of learning characters (step S31).

  In addition, the character information learning unit 114 adds the arrangement pattern and the partial character code of the identification source character to the learning list of the identification target character portion in the character shape element dictionary 14, and increments the number of learned characters (step S32).

  Thereafter, the character information learning unit 114 registers the character pattern data of the identification source character in the character recognition dictionary 12, and learns that the identification destination character is a candidate character (step S33).

DESCRIPTION OF SYMBOLS 1 Character identification device 2 External character file 3 External character form element storage file 4 Character code conversion definition list 5 Display part 6 Keyboard 11 Identification process part 12 Character recognition dictionary 13 OCR candidate character list 14 Character form element dictionary 15 radical candidate character list 16 Partial candidate Character list 17 Display candidate character list 18 Identification source / identification destination character correspondence list 111 OCR recognition unit 112 Candidate character list generation unit 113 Display candidate character list generation unit 114 Character information learning unit

Claims (10)

A character recognition storage unit for storing a dot pattern of characters;
About the character stored in the character recognition storage unit, an arrangement pattern indicating an arrangement of radicals, radical character information including a radical character code representing a radical, and a part representing a part other than the radical A glyph element storage for storing glyph element information of a part including a character code;
An external character storage unit for storing a dot pattern of an external character that is a character that is not included in a standardized character represented by a character code representing a predetermined character;
About the external character stored in the external character storage unit, an arrangement pattern indicating the arrangement of the radical, radical character information including a radical character code indicating the radical, and a partial character indicating a part other than the radical An external character-shaped element storage for storing character-shaped element information of a part including a code;
For the external character to be processed selected from the external character storage unit, the character recognition storage based on the external character dot pattern stored in the external character storage unit and the character dot pattern stored in the character recognition storage unit An OCR recognition unit that extracts a first candidate character that identifies an external character to be processed from characters stored in the unit;
For the external character to be processed, the radical character element information for the character stored in the character element storage unit and the radical character element information for the external character stored in the external character element storage unit Based on the above, a second candidate character for identifying the external character to be processed is extracted from the characters stored in the character-shaped element storage unit, and the external character to be processed is stored in the character-shaped element storage unit From the characters stored in the glyph element storage unit based on the glyph element information of the part for the character that has been performed and the glyph element information of the part of the external character stored in the external character shape element storage unit, A character identification device comprising: a candidate character list generation unit that extracts a third candidate character that identifies an external character to be processed. The character identification apparatus according to claim 1, wherein the character stored in the character recognition storage unit includes the standardized character, or the standardized character and the external character. The character identification device further includes:
The display candidate character list generation unit that assigns priorities according to the extent to which the candidate characters are included in the first to third candidate characters in an overlapping manner. The character identification device described in 1. The character information learning part which determines the identification destination character which is a character which identifies the external character which is the said process target from the said 1st candidate character thru | or a 3rd candidate character is provided. Character identification device. The character information learning unit, when there is a character common to each of the first candidate character to the third candidate character, determines the common character as the identification target character. Item 5. The character identification device according to Item 4. The character identification device further includes:
An output unit for outputting the first to third candidate characters;
The character identification device according to claim 4, wherein the character information learning unit determines the identification destination character based on a selection input designating the candidate character output to the output unit. 5. The character according to claim 4, wherein the character information learning unit generates a character correspondence list associating the identification source character and the identification destination character with the external character to be processed as the identification source character. Identification device. The character information learning unit uses the external character to be processed as an identification source character, the radical shape element information of the radical for the identification source character or the character shape element information of the portion, and the learning element information for the identification destination character The character identification device according to claim 4, wherein the character identification element information is added to the radical shape element information of the radical or the character shape element information of the portion of the identification target character. The candidate character list generation unit extracts the second candidate character and the third candidate character based on the added learning element information of the radical and the learning element information of the portion. The character identification device according to claim 8. For characters, an arrangement pattern indicating the arrangement of radicals, radical character element information including radical character codes representing radicals, and character element information of parts including partial character codes representing parts other than the radicals Is stored in the glyph element storage unit,
A process of storing the dot pattern of the character stored in the character-shaped element storage unit in a character recognition storage unit;
Processing for storing a dot pattern of an external character that is a character not included in the standardized character represented by a character code representing a predetermined character in the external character storage unit;
About the external character stored in the external character storage unit, an arrangement pattern indicating the arrangement of the radical, radical character information including a radical character code indicating the radical, and a partial character indicating a part other than the radical Processing to store the glyph element information of the portion including the code in the external character element storage unit;
For the external character to be processed selected from the external character storage unit, the character form element storage is performed based on the external character dot pattern stored in the external character storage unit and the character dot pattern stored in the character recognition storage unit. A process of extracting a first candidate character for identifying an external character to be processed from characters stored in a section;
For the external character to be processed, the radical character element information for the character stored in the character element storage unit and the radical character element information for the external character stored in the external character element storage unit Based on the above, a second candidate character for identifying the external character to be processed is extracted from the characters stored in the character-shaped element storage unit, and the external character to be processed is stored in the character-shaped element storage unit From the characters stored in the glyph element storage unit based on the glyph element information of the part for the character that has been performed and the glyph element information of the part of the external character stored in the external character shape element storage unit, A character identification method, comprising: causing a computer to execute a process of extracting a third candidate character that identifies an external character to be processed.

Images