JPH06290298A - Correcting method for erroneously written character - Google Patents

Abstract

PURPOSE:To correct an erroneously written character by using a correction pattern prepared by an operator himself and a co-operator by automatically correcting the erroneously written character by using cooccurence relation generated by his own computer and the correction pattern generated by the other computer. CONSTITUTION:Concerning data read from a word lattice file 303, a text correction aid program 301 performs both the automatic correction of the erroneously written character by using the cooccurence relation read from a cooccurence relation file 305 and an intermediate file 309 and the correction of the erroneously written character manually worked by the user and outputs the corrected result to a text file 307. When the user selects the character recognized text to be the object of correction, the erroneously written character in a correcting image is automatically corrected by using the cooccurence relation of the word stored in advance. When any erroneously written character exists in the character recognized text to be the object of correction, the correct character candidates of all the words are displayed, the inputs of both the word including the right character and the cooccurence word are accepted, and the new cooccurence relation is generated and stored from the word including the correct character inputted by the user and the cooccurence word.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION The present invention relates to a method for correcting typographical errors in text written in natural language. In particular,
Regarding correction of typographical errors included in character recognition results.

[0002]

2. Description of the Related Art Japanese Patent Laid-Open No. 4-323787 is an example of a conventional technique for correcting a character recognition result. This is a patent for a method of correcting misread characters. The present invention provides a co-occurrence word obtained by referring to a co-occurrence table in which words having a co-occurrence relationship with a certain word are arranged in order of frequency, and a candidate word or a single-character word having a word length of 2 or more and including candidate omissions. By collating with the candidate word of, the misread character is efficiently corrected and the misread correction rate is improved.

Japanese Patent Laid-Open No. 4-248640 can be cited as an example of a conventional technique for document processing in a distributed computing environment. This is a patent for a distributed processing device having a document logical structure. The present invention relates to a distributed processing apparatus for document logic, which updates, refers to, adds, or deletes entities based on a structural medium that does not depend on a specific model for the logical structure and attribute information of a document, and co-writes in document units. The purpose is to enable co-creation.

[0004]

SUMMARY OF THE INVENTION The present invention provides a means for a plurality of persons to jointly correct a typographical error included in the result of character recognition. In particular, it divides one document that has been character-recognized,
Providing a means to correct typos shared by multiple people.

[0005]

[Means for Solving the Problems] As means for correcting a typographical error included in a character recognition result, when a user selects a character-recognized text to be corrected, a co-occurrence relation of words stored in advance is used. If a typographical error in the correction target is corrected automatically, and if there is a typographical error in the recognized text that is the correction target, the orthographic candidates for all words are displayed, and the words containing the orthographic characters and their co-occurrence words are input. A method is used in which a new co-occurrence relation is generated and stored from a word including an orthography input by the user and its co-occurrence word.

Further, as means for jointly correcting a typographical error by a plurality of people, a typographical error is corrected automatically by using a co-occurrence relation stored in advance and a co-occurrence relation of words sent from another computer. A new co-occurrence relation generated by the own computer is stored in the own computer and distributed to other computers. Furthermore, as a means to correct the typographical error by dividing the same document and sharing it among multiple people, the co-occurrence relationship between the character-recognized text to be corrected and the word generated from the same text is used to correct the typographical error. A method is used in which automatic correction is performed and a computer name and a file name having the character-recognized text to be corrected are added to the newly generated co-occurrence relation.

In addition to this, as a means for correcting a typographical error included in the character recognition result, when the user selects the text for which the character is recognized as the correction target, the typographical error in the correction target is stored using the correctness relation stored in advance. Automatic correction of
In addition, if there is a typographical error in the character-recognized text that is the correction target, the orthographic candidates of all words are displayed, the input of the word including the orthographic characters is accepted, and the new correct or incorrect relation is input from the word including the orthographic characters input by the user. To generate and store. Furthermore, as a method for joint correction of typographical errors by multiple people, the typographical error in the correction target is automatically corrected by using the correctness relationship stored in advance and the correctness relationship of words sent from another computer. A means for storing a new correct / wrong relationship generated by a computer in its own computer and distributing it to another computer is used. Furthermore, as a means of dividing the same document and correcting the typographical error by sharing it among multiple people, the correctness of the word generated from the same text as the character-recognized text to be corrected is used to correct the erroneous character. A means for correcting the typographical error and adding the computer name and the file name having the character-recognized text to be corrected to the newly created correctness relationship is used.

[0008]

When the user selects the character-recognized text, the typographical error in the text is corrected by automatically correcting the typographical error in the correction target using the co-occurrence relation of the words stored in advance. If there are typographical errors in the text that has been recognized, the candidates for orthographic characters of all the words are displayed, the words that include the orthographic characters and their co-occurrence words are accepted, and the words that include the orthographic characters input by the user and their co-occurrence words are received. By generating a new co-occurrence relation from a word and storing it, a co-occurrence relation that has not appeared so far is generated and stored.

By automatically correcting typos using the co-occurrence relations stored in advance and the co-occurrence relations of words sent from another computer, the past co-occurrence relations appearing on the self-computer and the other computer can be corrected. Use to correct typographical errors in the text. By storing the new co-occurrence relation generated in the self-computer in the self-computer and distributing it to the other computer, the co-occurrence relation which has not appeared in the self-computer and the other computer until now is stored. By automatically correcting typographical errors using the co-occurrence relationship of words generated from the same text as the correction target, the current co-occurrence relationship between the self-calculator and other calculators that is currently being corrected is corrected. Correct typographical errors in a document using the co-occurrence relationships generated from the document.

By adding the computer name and the file name of the file having the text to be corrected to the newly generated co-occurrence relation, the co-occurrence relation that has not appeared until now with respect to the text currently being corrected can be exchanged with the other computer. Store in the calculator. When the user selects the character-recognized text, the typographical error in the text is corrected by automatically correcting the typographical error in the correction target using the correctness relation of the words stored in advance.

If there is a typographical error in the character-recognized text, the orthographical candidates of all the words are displayed, the input of the word including the orthographical characters is accepted, and a new correctness relation is input from the word including the orthographical characters input by the user. By generating and storing, a correctness relationship that has not appeared so far is generated and stored. By automatically correcting typographical errors using the correctness relations stored in advance and the correctness relations of words sent from other computers, the typographical errors in the text are made using the past correctness relations that have appeared on your computer and other computers. To fix.

By storing the new correct / wrong relationship generated by the own computer in the own computer and distributing it to the other computer, the correct / wrong relationship which has never appeared in the own computer and the other computer is stored. By automatically correcting typos using the correctness relationship between words generated from the same text as the correction target,
The typographical error in the document is corrected by using the correctness relation generated from the document currently being corrected among the past correctness relations appearing on the own computer and other computers.

By adding the computer name and the file name of the file having the text to be corrected to the newly created correctness relationship, the correctness relationship that has not appeared so far for the text currently being modified can be transmitted to the own computer and other computers. Remember.

[0014]

Embodiment An embodiment of the present invention will be described with reference to the drawings. Here, a Japanese text correction support system using the present invention will be described. FIG. 1 is a system flow. This system flow is for the text correction support program 301. Before describing this system flow, the hardware configuration and system configuration will be described with reference to FIGS. 2 and 3.

FIG. 2 shows the hardware configuration. The hardware configuration is such that a plurality of computers are connected by a communication line 213. Instructions and data can be exchanged between these computers. The computer is the main memory 201 (221), CP
U203 (223), CRT205 (225), secondary storage device 207 (227), keyboard 209 (22)
9) and the mouse 211 (231).

FIG. 3 shows the system configuration. The text correction support program 301 uses the word lattice file 303.
The co-occurrence relation file 30 for the data read from
5 and erroneous characters are automatically corrected using the co-occurrence relationship read from the intermediate file 309 and the intermediate file 309, and the correction result is output to the text file 307. The text correction support program 301 and the communication program 311 exist on the main storage device 201 (221). In the case of a computer that realizes virtual memory, it exists on the virtual memory.

The word lattice file 303, the co-occurrence relation file 305, the text file 307, and the intermediate file 309 exist on the secondary storage device 207 (227).

The programs of the text correction support program 301 and the communication program 311 are separate processes.
These two programs can exchange instructions and data by interprocess communication. The word lattice file 303 is a file that stores the word lattice of the character recognition result. The word lattice is obtained by dividing a certain character recognition result into a plurality of word divisions and prioritizing the division results. The word lattice file 303 will be described with reference to FIG.

FIG. 4 shows the logical structure of the word lattice file 303. The data 401 (410) is a block identifier. The block identifier is for identifying the character recognition unit. The data 403 (413) is a data identifier. The data 403 includes a host identifier 403a and a file identifier 403b for identifying the computer.
Data 404 (405, 406, 407, 408, 40)
9) is word information. The word information is the word headline 404a.
And part-of-speech information 404b and priority 404c.

Block identifiers are pointers 402 to data identifiers and pointers to word information (414, 415, 4).
16) and a pointer 417 to the next block identifier.

The word information is a pointer to the next word (41
8, 419, 420, 421, 422, 423, 42
Have 4). The co-occurrence relation file 305 is a file that stores a co-occurrence relation that is a correction pattern of a typographical error. The co-occurrence relation file 305 will be described with reference to FIG.

FIG. 5 shows the logical structure of the co-occurrence relation file 305. Data 501 is a co-occurrence relationship heading. The co-occurrence relationship heading is headword 501a and host identifier 5
01b and a file identifier 501c. The entry word 501a is a word including a typographical error. The data 503 is a correction word. The correction word 503 is the headword 501
It is a word that includes a proper letter for the typographical error included in a. The data 505 is co-occurrence data. The co-occurrence data includes co-occurrence word headline 505a, part-of-speech information 505b, and position information 50.
It consists of 5c. The co-occurrence word headline 505a is the headword 501
It is a word that co-occurs with a.

The co-occurrence relation heading is a pointer 5 to a correction word.
02 and a pointer 507 to the next co-occurrence relation. The correction word 503 has a pointer 504 to the co-occurrence data and a pointer 508 to the next correction word. The co-occurrence data 505 has a pointer 506 to the next co-occurrence data. The text file 307 is a file that stores the correction result of a typographical error.
The text file 307 will be described with reference to FIG.

FIG. 6 shows the logical structure of the text file. Field 601 is a block identifier field. The block identifier 401 of the word lattice file is stored in the block identifier field. Field 60
2 is a host identifier field. In the host identifier field, the host identifier 403 of the word lattice file
Store a. Field 603 is a file identifier field. The block identifier 403b of the word lattice file is stored in the file identifier field. Field 604 is a text field. The text field stores the word string corrected using the co-occurrence relationship. The intermediate file 309 is used to exchange the co-occurrence relationship between the text correction support program 301 and the communication program 311. About the intermediate file 309 FIG.
Will be explained.

FIG. 7 shows the logical structure of the intermediate file 309. The intermediate file 309 contains fields 701 and 70.
A record 710 having 2, 703, 704 is stored. The field 701 is an input / output flag field. The input / output flag field stores a flag for distinguishing between data to be distributed to an external computer and data acquired from the external computer. Field 702 is a host identifier field. The value of the host identifier 403a in the word lattice is stored in the host identifier field.

Field 703 is a file identifier field. The file identifier field stores the value of the file identifier 403b in the word lattice. Field 704 is a data field. The data field stores the co-occurrence relationship. The co-occurrence relation stored in the data field is equal to the logical structure of the co-occurrence relation file 305 excluding the host identifier 501b and the file identifier 501c.

The communication program 311 performs a process of distributing the data of its own computer to another computer and a process of fetching the data of the other computer into its own computer. The processing of the communication program 311 will be described with reference to FIG.

FIG. 8 is a processing flow of the communication program 311. A process 801 is a process of waiting for message input from the own computer and other computers. A process 803 is a process of determining the input destination computer of the input obtained in the process 801. If the input destination is the own computer, the co-occurrence relation with the input / output flag "OUT" is distributed from the intermediate file 309 to the other computer, and if the input destination is another computer, the input / output flag "I" is added to the distributed co-occurrence relation.
N is added and stored in the intermediate file 309. Process 80
5 indicates that the input / output flag in the intermediate file 309 is "OUT"
Is a process of determining the presence or absence of a co-occurrence relationship. OU
The record having the T "flag is created by the co-occurrence relation distribution processing 1407 of the text correction support program 301. If the corresponding co-occurrence relation exists, the co-occurrence relation is distributed to another computer and the distributed co-occurrence relation is distributed. Relationship intermediate file 30
Delete from 9. If the corresponding co-occurrence relation does not exist, the process returns to processing 801.

Process 807 is a process for distributing the co-occurrence relation to other computers. The process 809 is a process of deleting the co-occurrence relation distributed to another computer from the intermediate file 309. The process 811 is a process of adding the input / output flag “IN” to the distributed co-occurrence relation and storing it in the intermediate file 309.

FIG. 1 shows a text correction support program 301.
Is the system flow of. The text correction support program 301 reads the word lattice to be corrected from the word lattice file 303, supports the automatic correction of the word string using the co-occurrence relation, the manual correction of the word string, and outputs the correction result to the text file 307. It is a program to do.
Process 101 is a process for displaying a screen. The screen displayed in the process 101 is shown in FIG.

FIG. 9 is a screen for displaying a word string read from the word lattice file 303. The display frame 903 is a frame for displaying a word string. The display frame 903 is a character recognition unit display frame 903a, 903b, 903c, 903d, 9
It consists of 03e. To select the character recognition unit that the user wants to correct, the character recognition unit display frames 903a, 903b, 9
03c, 903d, 903e is the user's mouse 211
Instruct using (231).

Symbols 905, 907, 909, 911, 9
Reference numerals 13, 914, 915, and 916 are text correction support program commands. Symbol 905 is a program end command. Symbol 907 is word lattice file 3
It is a command to read the word lattice from 03. Symbol 909 is a command for storing the corrected word string in the text file 307. Reference numeral 911 is a manual text correction start command. Symbols 913, 914, 91
Reference numerals 5 and 916 are commands for scrolling the word string displayed in the frame 903. These commands are mouse 21
It is activated by instructing using 1 (231).

Process 103 is a process of waiting for an input from the user. Process 105 is a process of determining the input obtained in process 103. When the program end command 905 is obtained in the process 103, the process 107 is executed. Process 103
If you get the command 907 to read the word lattice with
The process 109 is executed. When the command 909 for storing the word string corrected in the process 103 is obtained, the process 111 is executed. In process 103, scroll commands 913, 91
When 4, 915 and 916 are obtained, the processing 117 is executed. When the character recognition unit selection is obtained in process 103, process 1
Execute 13. When the text correction start command 911 is obtained in the process 103, the process 115 is executed. Processing 10
7 is a process for deleting the screen displayed in the process 101.
The process 109 is a process of reading a word lattice from the word lattice file 303. The process 109 will be described with reference to FIG.

FIG. 10 is a processing flow of processing for reading the word lattice. Process 1001 is a process of acquiring the file name of the word lattice file 303, the start position of the word lattice, and the number of word lattices from the user. Processing 1
A process 003 is a process for opening the file designated in the process 1001. A process 1005 is a process success / failure determination process of the process 1003. If the process 1003 is successful, the word lattice file is read. Processing 100
If the process of 3 fails, the process of FIG. 10 ends.
Process 1007 is a process of skipping the word lattice up to the start position of the word lattice obtained in process 1001. A process 1009 is a process of reading word lattices for the number of cases obtained in the process 1001 and displaying word information with the highest priority in the display frame 903. Process 1011 is process 10
This is the process of closing the file opened in 03. Process 111
Is a process of storing the correction result in the text file 307. The process 111 will be described with reference to FIG. FIG. 11 is a processing flow of processing for storing the correction result in the text file.

Process 1101 is a process for acquiring the file name of the text file 307 from the user. Process 11
Reference numeral 03 is a process for opening the file designated in the process 1101. Processing 1105 is processing success / failure determination processing of processing 1103. If the process 1103 is successful,
Output the correction result to a text file. Process 1103
If the process of 1 fails, the process of FIG. 11 ends. The process 1107 is a process of outputting the block identifier, the host identifier, the file identifier, and the word information with the highest priority to the file designated in the process 1103. Processing 1109 is processing for closing the file opened in processing 1103. Process 113 is a process of correcting a typo using the co-occurrence relation for the character recognition unit selected in process 103. The process 113 will be described with reference to FIG.

FIG. 12 is a processing flow of processing for automatically correcting a typographical error using the co-occurrence relationship. A process 1201 is a counter initialization process for counting the number of pieces of word information having the highest priority (hereinafter, referred to as a first candidate). Process 120
Reference numeral 3 is an acquisition process of the number of words to be matched. The matching target is the first candidate. A process 1205 is a termination determination of the correction process. When the word number counter exceeds the number of words to be matched, the process of FIG. 12 ends. When the word number counter does not reach the number of words to be matched, word correction processing using the co-occurrence relation is performed. The process 1207 is the co-occurrence relation file 305.
Is a process of matching the first candidate of the i-th word. In the matching, the entry word 501a of the co-occurrence relation file 305 is equal to the first candidate and the host identifier 501b is the host identifier 4
03a and the file identifier 501c is equal to the file identifier 403b, and the co-occurrence word heading 505 of the co-occurrence data 505 possessed by the retrieved co-occurrence relationship is searched.
a, part of speech information 505b, position information 505c and all first
Match the candidates.

Process 1209 is a success / failure determination process of process 1207. If the process 1207 is successful,
The first candidate of the i-th word is set as the correction word included in the co-occurrence relation. If the process 1207 fails, the intermediate file 3
Matching is performed using the co-occurrence relation in 09. The process 1211 is the correction word included in the co-occurrence relation checked in the process 1207 is i
This is a process of substituting the first candidate for the second word. Process 12
13 is a process of collating the co-occurrence relation of the record in which the input / output flag field 701 in the intermediate file 309 is "IN" with the first candidate of the i-th word. The collation searches for a co-occurrence relation in which the entry word 501a included in the data field 704 of the intermediate file 309 is equal to the first candidate, the host identifier field 702 is equal to the host identifier 403a, and the file identifier field 703 is equal to the file identifier 403b. The co-occurrence word headline 505a and the part-of-speech information 50 of the co-occurrence data 505 possessed by the retrieved co-occurrence relationship
5b, position information 505c and all the first candidates are matched.

The process 1215 is a process of storing the co-occurrence relation of the record in which the input / output flag field 701 in the intermediate file 309 is "IN" in the co-occurrence relation file 305. The stored record is deleted from the intermediate file 309. Process 1217 is a success / failure determination process of process 1213. If the processing 1213 is successful, the processing 1
Perform 211. If the process 1213 is unsuccessful, the co-occurrence relation is collated for the next first candidate. Process 1219 is a process of counting the number of first candidates that have been collated. Process 115 is a manual correction process by the user of the word string. The process 115 will be described with reference to FIG.

FIG. 14 is a processing flow of the manual correction processing by the user. A process 1401 is a process of displaying a screen for manual correction. The screen displayed in the process 1401 is shown in FIG.
Will be explained.

FIG. 13 shows a screen for manual correction. Symbol 1
Reference numerals 303, 1304, 1305, 1306, and 1307 are commands for manual correction. Reference numeral 1303 is an end command. Symbols 1304, 1305, 1306, 130
Reference numeral 7 is a scroll command for the word display frame 1302.
The display frame 1302 is a frame for displaying word information. Symbol 1
310, 1311, 1312, 1313, and 1314 are the first candidate word information. Symbols 1315, 1316, 1
Word information 317, 1318, and 1319 are word information having a lower priority than the word information of the first candidate (hereinafter referred to as the nth candidate). The first candidate for the nth candidates 1315, 1316 is symbol 1310, the first candidate for the nth candidate 1317 is symbol 1312, and the first candidate for the nth candidates 1318, 1319 is symbol 1313.

Process 1403 is a process for correcting a word by the user. The process 1403 will be described with reference to FIG. FIG. 15 is a processing flow of word correction processing by the user. A process 1501 is waiting for an input from the user. Process 1503 is the input determination process obtained in process 1501. If a scroll command is input in processing 1501, the display frame 1302 is scrolled. Process 1501
If the nth candidate symbol is selected at, the typographical error is corrected and the co-occurrence relation is created. If the end command is input in processing 1501, the processing of FIG. 15 is ended. A process 1505 is a process of scrolling the display frame 1302. Display frame 130
2 selects the symbol 1304 to move up, the symbol 1305 selects the symbol down, and the symbol 1306 selects the symbol to the left, the symbol 1
Selecting 307 scrolls to the right.

The process 1507 is a process for creating a co-occurrence relation with the first candidate as the index for the selected nth candidate.
Process 1509 is a process of substituting the selected n-th candidate for the co-occurrence relation correction word created in process 1507. Processing 1
511 is a display frame 1 in which the selected nth candidate is the first candidate.
This is processing to be displayed on 302. The process 1513 is waiting for the input of the co-occurrence word. The user inputs the co-occurrence word with the mouse 221.
This is done by instructing the word information using (231). Process 1515 is the input determination process obtained in process 1513. If the symbol 1303 is input in process 1513, the process returns to process 1501. If a scroll command is input in processing 1513, the display frame 1302 is scrolled. When the word information is input in the process 1513, the input word information is used as the co-occurrence data of the co-occurrence relation created in the process 1507.

Process 1517 is a process of storing the word information input in process 1513 as co-occurrence data of the co-occurrence relationship created in process 1507. Process 1405 is process 14
This is a process of storing the co-occurrence relation created in 03 in the co-occurrence relation file 305. Process 1407 is a process of distributing the co-occurrence relationship created in process 1403 to other computers. In the process 1407, “OUT” is set in the input / output flag field 701 of the intermediate file 309, the host identifier 403a of the co-occurrence relation is set in the host identifier field 702, and the host identifier 40 of the co-occurrence relation is set in the file identifier field 703.
3b stores the record having the co-occurrence relationship created in the data field 704, and sends a message to the communication program 311.

Process 1409 is a process for deleting the screen displayed in process 1401. Process 117 is a scroll process of the display frame 903. When the symbol 913 is selected, the display frame 903 moves up; when the symbol 914 is selected, it moves down;
Selecting 5 scrolls to the left and selecting symbol 916 scrolls to the right.

The effect of this embodiment is that the co-occurrence relation created by the self-computer is distributed to the other computers, so that the co-occurrence relation of the self-computer also exists on the other computers. As a result, even if the own computer is stopped, other computers can correct typographical errors using the co-occurrence relation of the own computer.

As another embodiment, a method of correcting a typographical error using the correctness relationship shown in FIG. 17 instead of the cooccurrence relationship will be described. The correctness relationship will be described with reference to FIG. The correctness relationship is stored in the correctness relationship file 1605. Data 1
Reference numeral 701 is a headline of the correctness relationship. The headline of the correctness relationship includes a headword 1701a, a host identifier 1701b, and a file identifier 1701c. Entry word 170
1a is a word containing a typographical error. Data 1703
Is a modified word. The modified word 1703 is a headword 1701a
It is a word that contains orthographic characters for typographical errors included in. The headline of the correctness relationship is the pointer 1702 to the correction word.
And a pointer 1705 to the next correctness relationship. The correction word 1703 has a pointer 1704 to the next correction word.

The system configuration is as shown in FIG. The text correction support program 1601 automatically corrects typographical errors in the data read from the word lattice file 303 using the correctness / correlation relationship read from the correctness / correlation file 1605 and the intermediate file 1609, and corrects the typographical errors by the user's manual work. , And outputs the correction result to the text file 307. Text correction support program 16
01 and the communication program 1611 are stored in the main storage device 201 (2
21) present above. In the case of a computer that realizes virtual memory, it exists on the virtual memory.

The word lattice file 303, the correct / wrong relationship file 1605, the text file 307, and the intermediate file 1609 exist on the secondary storage device 207 (227). The text correction support program 1601 and the communication program 1611 are different processes.
These two programs can exchange instructions and data by interprocess communication.

FIG. 18 shows the logical structure of the intermediate file 1609. The intermediate file 1609 has a field 180
Record 18 with 1, 1802, 1803, 1804
10 is stored. Field 1801 is an input / output flag field. The input / output flag field stores a flag for distinguishing between data to be distributed to an external computer and data acquired from the external computer. Field 180
2 is a host identifier field. The value of the host identifier 403a in the word lattice is stored in the host identifier field. Field 1803 is a file identifier field. The file identifier field stores the value of the file identifier 403b in the word lattice. Field 1804 is a data field. The data field stores the correctness relationship. The true / false relationship stored in the data field is determined from the logical structure of the true / false relationship file 1605 by the host identifier 1701b and the file identifier 170.
Equivalent to 1c.

FIG. 19 is a processing flow of the communication program 1611. Processing 1901 is processing for waiting for message input from the own computer and other computers. Process 1903 is process 1
This is the determination processing of the input destination computer of the input obtained in 901. If the input destination is the own computer, the correctness relationship with the input / output flag "OUT" is distributed from the intermediate file 1609 to the other computer, and if the input destination is another computer, the input / output flag "IN" is attached to the distributed correctness relationship. It is stored in the intermediate file 1609. A process 1905 is a process of determining the presence / absence of a correct / incorrect relationship in which the input / output flag in the intermediate file 1609 is “OUT”. The creation of the record having the “OUT” flag is performed by the correctness / correlation distribution process 1407 of the text correction support program 1601. If there is a corresponding right / wrong relationship, the right / wrong relationship is distributed to other computers, and the distributed right / wrong relationship is deleted from the intermediate file 1609. If there is no corresponding right / wrong relationship, the process returns to processing 1901.

Process 1907 is a process for distributing the correctness relationship to other computers. A process 1909 is a process for deleting the correct / wrong relationship distributed to another computer from the intermediate file 1609. A process 1911 is a process of adding the input / output flag “IN” to the distributed correctness relationship and storing it in the intermediate file 1609. The system flow of the text correction support program 1601 will be described with reference to FIG. Process 2001 is a process of correcting a typographical error by using the correctness relationship with the character recognition unit selected in process 103. The process 2001 will be described with reference to FIG.

FIG. 21 is a processing flow of processing for automatically correcting a typographical error by using the correctness relationship. Process 2101 is a counter initialization process for counting the number of first candidates. Process 2103 is a process of acquiring the number of words to be matched.
The matching target is the first candidate. A process 2105 is a determination of the end of the correction process. When the word number counter exceeds the number of words to be collated, the processing of FIG. 21 ends. When the word number counter does not reach the number of words to be matched, the word correction process using the correctness relationship is performed. The process 2107 is a process of collating the correct / wrong relationship file 1605 with the first candidate of the i-th word. The collation is headword 16 of the true / false relationship file 1605.
01a is equal to the first candidate and host identifier 1601b
Is equal to the host identifier 403a and the file identifier 1
601c is to search for a true / false relationship equal to the file identifier 403b.

Process 2109 is a success / failure determination process of process 2107. If the process 2107 is successful,
The first candidate of the i-th word is set as the correction word included in the correctness relationship. If the process 2107 fails, the intermediate file 1
Matching is performed using the correctness relationship in 609. Process 2111
Is a process of substituting the correction word included in the correctness relation checked in the process 2107 into the first candidate of the i-th word. Process 2
Reference numeral 113 denotes a correct / wrong relationship of records in which the input / output flag field 1801 in the intermediate file 1609 is “IN” and i
This is a process of collating the first candidate of the second word. The collation is performed by searching for a correct relationship in which the entry word 1701a included in the data field 1804 of the intermediate file 1609 is equal to the first candidate, the host identifier field 1802 is equal to the host identifier 403a, and the file identifier field 1803 is equal to the file identifier 403b. is there.

The process 2115 is a process for storing the correct / wrong relationship of the record of which the input / output flag field 1801 in the intermediate file 1609 is “IN” in the correct / wrong relationship file 1605. The stored record is the intermediate file 1609.
Remove from. Process 2117 is a success / failure determination process of process 2113. If the process 2113 is successful, the process 2111 is performed. If the process 2113 fails, the correctness of the next first candidate is checked. Process 2119 is a process of counting the number of first candidates checked. The process 2002 is a manual correction process by the user of the word string. The process 2002 will be described with reference to FIG. Process 2203 is a word correction process by the user. The process 2203 will be described with reference to FIG.

FIG. 23 is a processing flow of word correction processing by the user. A process 2301 is waiting for an input from the user. The process 2303 is the input determination process obtained in the process 2301. If a scroll command is input in processing 2301, the display frame 1302 is scrolled.
If the nth candidate symbol is selected in the process 2301, the typographical error is corrected and the correctness relation is generated. If the end command is input in process 2301, the process of FIG. 23 is ended. A process 2307 is a process of forming a correct / incorrect relationship using the first candidate as a heading for the selected nth candidate. Process 2309 is a process of substituting the selected n-th candidate for the corrective word of the correctness relationship created in process 2307.

The process 2311 sets the selected nth candidate to the first candidate.
This is a process of displaying in the display frame 1302 as a candidate. Process 2317 is a process of storing the correctness relationship created in process 2307. A process 2205 is a process of storing the correct / wrong relationship created in the process 2203 in the correct / wrong relationship file 1605. Process 2207 is a process of distributing the correctness relationship created in process 2203 to other computers. In process 2207, the input / output flag field 180 is added to the intermediate file 1609.
“1” is “OUT”, the host identifier field 1802 is the host identifier 403a having the correct relationship, and the file identifier field 1803 is the host identifier 403b having the correct relationship.
Is stored in the data field 1804, and the record having the correct relationship is stored, and the message is sent to the communication program 1611. The effect of this embodiment is that the typographical error can be corrected by a simpler process than that using the co-occurrence relation.

[0057]

The effect of the present invention is to automatically correct a typographical error by using the co-occurrence relation generated by the self-computer and the correction pattern generated by another computer, so that the correction pattern created by myself and the collaborator can be obtained. Since the typographical error can be corrected by using, the character recognition result with few typographical errors can be obtained.

[Brief description of drawings]

FIG. 1 is a diagram illustrating a system flow.

FIG. 2 is a diagram illustrating a hardware configuration.

FIG. 3 is a diagram illustrating a system configuration.

FIG. 4 is a diagram illustrating a logical structure of a word lattice file.

FIG. 5 is a diagram illustrating a logical structure of a co-occurrence relation file.

FIG. 6 is a diagram illustrating a logical structure of a text file.

FIG. 7 is a diagram illustrating a logical structure of an intermediate file.

FIG. 8 is a diagram illustrating a processing flow of a communication program.

FIG. 9 is a diagram illustrating a text display screen.

FIG. 10 is a diagram illustrating a processing flow of processing for reading a word lattice from a word lattice file.

FIG. 11 is a diagram illustrating a processing flow of processing for storing a correction result in a text file.

FIG. 12 is a diagram illustrating a processing flow of processing for automatically correcting a typographical error using a co-occurrence relationship.

FIG. 13 is a diagram illustrating a screen for manual correction.

FIG. 14 is a diagram illustrating a processing flow of manual correction processing by a user.

FIG. 15 is a diagram illustrating a processing flow of word correction processing by a user.

FIG. 16 is a diagram illustrating a system configuration of another embodiment.

FIG. 17 is a diagram illustrating a correctness relationship of another embodiment.

FIG. 18 is a diagram illustrating a logical structure of an intermediate file according to another embodiment.

FIG. 19 is a diagram illustrating a communication program according to another embodiment.

FIG. 20 is a diagram illustrating a system flow of another embodiment.

FIG. 21 is a diagram illustrating an automatic erroneous character correction process according to another embodiment.

FIG. 22 is a diagram illustrating a processing flow of manually correcting a typographical error in another embodiment.

FIG. 23 is a diagram illustrating a processing flow of word manual correction according to another embodiment.

[Explanation of symbols]

113 ... Automatic word string correction processing. 115 ... Manual correction processing of word string. 301 ... Text correction support program. 303 ...
Word lattice file. 305 ... Co-occurrence relation file. Three
07 ... Text file. 309 ... Intermediate file. 31
1 ... communication program.

Claims (6)

[Claims] 1. A method for correcting a typographical error included in a character recognition result, wherein when a user selects a character-recognized text to be corrected, the typographical error is automatically detected using a co-occurrence relation of words stored in advance. If there is a typographical error in the recognized text that has been corrected and the character to be corrected is displayed, the orthographic candidates for all words are displayed, and the word containing the orthographic characters and its co-occurrence word are accepted and entered by the user. A method for correcting a typographical error, which is characterized in that a new co-occurrence relation is generated and stored from a word including an orthographic character and its co-occurrence word. 2. The correction method according to claim 1, wherein the typographical error is corrected automatically by using the co-occurrence relation stored in advance and the co-occurrence relation of the words sent from another computer, and a new character generated by the own computer is generated. A method of correcting a typographical error, which is characterized by storing the co-occurrence relationship in the own computer and distributing it to other computers. 3. The correction method according to claim 2, wherein erroneous characters are automatically corrected using a co-occurrence relation of words generated from the same text as the character-recognized text to be corrected,
A method of correcting a typographical error, which comprises adding a computer name and a file name having a character-recognized text to be corrected to a newly generated co-occurrence relation. 4. A method of correcting a typographical error included in a character recognition result, wherein when the user selects a character-recognized text to be corrected, the typographical error is automatically corrected using a pre-stored correctness relationship. In addition, if there is a typographical error in the character-recognized text to be corrected, all orthographic word candidates are displayed, and the input of the word containing the orthographic characters is accepted. A method of correcting a typographical error characterized by generating and storing a relationship. 5. The correction method according to claim 4, wherein a typographical error is corrected automatically by using a correctness relationship stored in advance and a correctness relationship between words sent from another computer, and a new correctness relationship generated by the own computer. A method of correcting a typographical error, which is characterized by storing in a computer and distributing it to other computers. 6. The correction method according to claim 5, wherein the typographical error is automatically corrected by using the correctness relationship between words generated from the same text as the character-recognized text to be corrected,
A method of correcting a typographical error, which comprises adding a computer name and a file name having a character-recognized text to be corrected to a newly created correctness relationship.