JP2011150436A - Method for substituting character data - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To substitute characters which may be erroneously read with correct characters. <P>SOLUTION: In a method for substituting character data, document data which is carried out character recognition by an OCR (optical character reader) is made to collate with the character data registered in a first substitution table and a second substitution table, and the similar word contained in the document is automatically substituted by a correct word or a mark identifiable from a similar character and output as data for editing. History for converting the similar characters and identifiable marks into the correct words can be accumulated in a history table of a data accumulation part, and can automatically registered as similar word data of the first substitution table. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a character that may be erroneously read in an operation of processing character data recognized by an OCR (optical character reader). It relates to a method of replacing ka and kanji “power”, katakana “ro” and kanji “mouth” and the like.

  Although digital information technology is now widespread, paper manuscripts are widely used as an information transmission medium. As a method for easily converting information obtained from a paper document into data, a method of recognizing character data with an OCR (optical character reader) and converting it into data is common. High recognition accuracy can be obtained by improving the reading resolution of OCR (optical character reader) and the development of image analysis technology. However, there is a possibility of erroneous reading (possibility of erroneous reading). Is still an error in identifying katakana “K” and kanji “power”, katakana “ro” and kanji “mouth”, etc.) Many people have to rely on manual modifications. In order to correct misread characters, temporarily convert them to “marks that can be recognized as misreadable characters”, then refer to the similar word list prepared in advance and manually enter the correct characters. I have been working on replacing it.

  As a method of performing high-precision correction on document data after recognition by an OCR (optical character reader) and reducing the burden of subsequent manual correction, the recognition result of the OCR (optical character reader) is used. There is a method of applying a morphological analysis, estimating a category to which a corresponding document belongs, and correcting a term error in the document by using a dictionary related to the estimated category preferentially (see Patent Document 1).

  To replace similar characters, perform a morphological analysis of Japanese kana-kana mixed text, divide it into units required for analysis, and use “-(sound drawing)”, “-(minus)”, “ There is a character replacement device (see Patent Document 2) that replaces “-(hyphen)”.

JP 2008-225695 A Japanese Patent No. 3407293

Character information provided to the world using a paper medium such as a book, booklet, or newspaper may not be confusing to the reader when the character information has legal effect or when reading the character information. In those cases, the text information must be highly accurate and should not be provided with typographical errors. Therefore, in the work of creating character data on the editing side, efforts must be made to eliminate as many typographical locations as possible. On the other hand, in an editing process such as processing and creation from character information of a manuscript on a paper medium to editing data, an OCR (optical character reader) is installed for the purpose of labor saving in manual editing that requires a lot of time. It is often used to read character information on a paper medium and process it into character data. Conventionally, when character data is processed from a paper document by an OCR (optical character reader), it may be erroneously processed to a character different from the character in the paper document. The text data obtained in step (3) is visually proofread on the editing screen. When a character that seemed to have been processed by mistake was found, it was replaced with a normal character by manual input. After these editing steps, in order to check the consistency between the created edit data and the original, the paper medium output as the edit data and the original are compared and visually compared. In the work, it is not a matter of comparing characters while reading the sentence and considering the meaning of the character or word, but only comparing the difference in character shape of each character, and includes typographical errors similar to the normal character. If the sentence is in a state, it is difficult to find a typographical place or an overlooked part in the editing process, and it takes a lot of time for the confirmation work.
The problem to be solved by the present invention is to reduce the time and work required for editing and proofreading that are performed manually, to correct a part that is not erroneous, or to overlook an erroneous part, etc. Is to prevent human error. For this purpose, a system that utilizes programming technology determines a misread character on the editing data, and automatically replaces the misread character with a correct character, for example. A method is desired.

  In Patent Literature 1, it is assumed that a category (classification / field) classification to be used and a dictionary related to the category can be generated based on an arbitrary classification of the creator. That is, the category and the dictionary related to the category are endless depending on the information to be handled. If it is used for proofreading a document for special business use with high accuracy, the utility value is increased by defining a special dictionary suitable for the field, but it is used for general purposes (newspapers, general books, etc.). The maintenance of the data requires an enormous amount of time and effort, and there are major problems in its maintenance and maintenance.

  In Japanese Patent Laid-Open No. 2004-260260, the conditions for efficiently maintaining the accuracy are limited to sentences having specific rules having at least a mathematical expression, an address, a telephone number, a kana word, a foreign language word, and the like. In general, a document submitted as a paper manuscript is not limited to this, and it is considered that the versatility for maintaining the accuracy of words including similar characters is low.

  SUMMARY OF THE INVENTION The present invention aims to solve the above-described problems, and there is a possibility of erroneous reading in an operation of processing character data of a paper document read by an OCR (optical character reading device) into editing data. It is to provide a method for replacing a certain character with a normal character.

  In a method for replacing character data by a character data replacement system including a data storage unit and a data generation unit, the data storage unit includes a first replacement table, a second replacement table, and a history table, and the data generation unit Is erroneously read into the data read by the optical character reading means, the optical character reading means for reading the character data acquired by the data acquisition means, the data acquisition means for acquiring the image data describing the characters In the first replacement means, a first replacement means for searching whether or not there is a word including a character that can be detected and replacing it with a word registered in the first replacement table if it exists, A character that may not be replaced is searched for a character that may be erroneously read, and is compared with a similar character list in the second replacement processing table to be erroneously read. Second replacement means for replacing with a mark that can be recognized as a possible character, replacing the data including the character replaced with the mark with a correct word, and storing the replaced data in the history table; In this method, character data is replaced by an editing data generation unit that generates the editing data.

  Time when a character that may be misread is detected and converted to a mark that can be recognized as a character that may be temporarily misread by program processing, and then the part is manually replaced When the character data replacement method according to the present invention is used, the time can be shortened by improving the efficiency of the editing work. Further, by considering convenience so that anyone can handle it, human errors in the editing process are prevented in advance, and edit data with higher reliability and accuracy than before can be obtained.

  For the creation and maintenance of similar word lists required for correction work, existing data and newly generated words of character string patterns including characters that may be misread are managed to generate newly generated words. It is necessary to grasp the frequency and occurrence pattern. However, since it takes a certain amount of time for a human to grasp and count the occurrence frequency and occurrence pattern, it has been difficult to update the data in a timely manner and maintain its accuracy. However, it is possible to count and process data at any time and timing by accumulating words in the string pattern including newly generated characters that may be misread and storing them in the history table each time. This makes it possible to maintain the accuracy of the list, thereby further improving the accuracy of the character data replacement processing.

It is a block diagram which shows the character data replacement system of this invention. It is a flowchart figure of the character data replacement method. It is a flowchart figure which shows the flow of embodiment of this invention. It is a figure which shows the similar word list | wrist of a 1st replacement table. It is a figure which shows the example of replacement when the word registered into the 1st replacement table is read. It is a figure which shows the preparation method of 1st replacement table data (1). It is a figure which shows the preparation method of the data (2) of a 1st replacement table. It is a figure which shows a similar character list. It is a figure which shows the example of replacement when the word which is not registered into the 1st replacement table is read.

  Embodiments of the present invention will be described with reference to the drawings. However, the present invention is not limited to the embodiments described below, and various other embodiments can be implemented within the scope of the technical idea described in the claims.

  A block diagram of the character data replacement system 1 of the present invention is shown in FIG. As shown in FIG. 1, the character data replacement system 1 includes a data generation unit 2 and a data storage unit 3. Furthermore, the character data replacement system 1 may include an update data generation unit as shown in a dashed line in FIG.

  As shown in FIG. 1, the data generation unit 2 includes a data acquisition unit 101, an optical character reading unit 102, a first replacement unit 103, a second replacement unit 104, and an edit data generation unit 105. The data acquisition unit 101 acquires image data 301 from a paper document using a scanner, a digital camera, or the like. There is no problem with electronic image data obtained from outside the company. The storage format of the image data 301 is not particularly limited, and is a format typified by JPEG, BMP, TIFF and the like, which are general image file storage formats. The image data 301 may be acquired from inside or outside the company via a network.

  The optical character reading unit 102 of the data generation unit 2 converts the image data 301 acquired by the data acquisition unit 101 into document data 302. The optical character reader 102 is not particularly limited as long as it is general optical character reader software (OCR software). The image data 301 acquired by the data acquisition unit 101 is preferably subjected to processing for removing unnecessary image information other than character information.

The first replacement unit 103 of the data generation unit 2 searches the optical character reading unit for whether there is a word including a character that may be erroneously read in the document data 302 read by the 102. If there is a word including a character that may be erroneously read, the word is replaced with similar word data registered in the first replacement table 201. FIG. 5 shows a processing example when a word including a character that may be erroneously read is registered in the first replacement table.
The word “population density” includes a letter “mouth” that may be misread. Two words of “population density” and “human density” are registered in the first replacement table, and if one of these words is read, the correct word It is automatically converted to "population density". A method of creating similar word data used in the first replacement table will be described with reference to FIG.
As a method of registering similar words in advance, there is a method in which many similar word lists can be created by paying attention to the fact that similar words can be assumed based on characteristics such as the arrangement of character types before and after similar characters. As a method,
(a) There are no words with katakana between kanji. For example, a word such as “gravity acceleration” in which the kanji “force” is “k” in katakana.
(b) There are no words with kanji between katakana and katakana. For example, the word “ta” in “digital” refers to the word “yu” in Chinese characters.
(c) There is no word that begins with the letter “n” after the kanji. For example, a word such as “protein” where “Ta” is the Chinese character “Yu”.
(d) There is no word that starts with the sound “tsu”, the roar “nya-yu-yo” or the sound-drawing “-” after the kanji. For example, the word “ni” in “new” is a word “ni”, or the word “ro” in “roma” is “mouth”.
(e) There are no words that contain similar characters in commonly used words. For example, the word “symptom” of “syndrome” is “症候群”.
As described above, it is possible to enrich similar word data by predicting a character string pattern of a word including a character that may be misread.
As another example, for example, in the case of the word “operator”, the katakana “pe” has three similar shapes: hiragana “pe”, katakana “be”, and hiragana “be”. There are characters (hereinafter referred to as “typical characters”) that are easy to read with OCR (optical character reader) such as “he-pe-be”, “na-me”, “n-so”. In the case of “-”, it is negatively categorized, and the katakana “ta” is similar to the kanji “Yu”, so there are a total of 15 types of character patterns for the word “operator”. By organizing the above, it is possible to further enhance the similar character list.

  The second replacement unit 104 of the data generation unit 2 searches for a character that may be erroneously read that has not been replaced by the first replacement unit 103, and collates it with a similar character list in the second replacement processing table 202, thereby causing erroneous reading. It is replaced with a mark that can be recognized as a character that may be used. In the embodiment for carrying out the invention, an example is shown in which “ゲ (getter mark)” is used as a mark that can be recognized as a character that may be misread. FIG. 9 shows an example of processing by the second replacement unit for characters that may not be replaced by the first replacement unit and may be misread. When a word that is not registered in the first replacement table 103 called “demographics” is read, the character “B” or “mouth” that may be erroneously read is temporarily displayed in the mark “〓” that can be recognized as it. Is replaced by The word including the character replaced with the “〓” mark can be replaced by selecting the correct word “population” from the dictionary table. Further, the data created in the second replacement means 104 is stored in the history table 204 of the data storage unit 3. A word including a character replaced with the “〓” mark can be converted into a correct word by manual input.

  The editing data generation unit 105 of the data generation unit 2 generates editing text data 304 from the data 303 subjected to the replacement processing by the first replacement unit 103 and the second replacement unit 104. The generated editing data 304 is preferably output after a spell check with commercially available software or the like.

  As shown in FIG. 1, the data storage unit 3 includes a first replacement table 201, a second replacement table 202, and a history table 203. The first replacement table 201, the second replacement table 202, and the history table 203 are stored in a general computer hard disk or the like.

  The first replacement table 201 of the data storage unit 3 holds similar word data, and the similar word data can be updated using the data accumulated in the history table 203 at an arbitrary period and timing.

  The second replacement table 202 of the data storage unit 3 holds a similar character list and a dictionary table 205 in which characters that may be misread are registered. In this embodiment, an example in which the dictionary table 205 is attached to the second replacement table 202 is shown, but the present invention is not limited to this.

  The history table 204 of the data storage unit 3 can store data subjected to replacement processing.

  As shown in FIG. 1, the update data generation unit 4 includes an update data creation unit 106 and a replacement table update unit 107.

  The update data creation means 106 of the update data generation unit 4 converts the data stored in the history table 204 into text data, and updates and processes the data.

  The replacement table update unit 107 of the update data generation unit 4 uses the data generated by the update data generation unit 106 to update the similar word data held in the first replacement table.

  A flowchart of the character data replacement method is shown in FIG.

  Image data in which characters are described is acquired by the data acquisition means F101 which is the first step.

  In the second step, the image data obtained in the first step is converted into document data (text data) recognized by the optical character reading means F102.

  The third step is the first replacement unit F103, which searches for data containing characters that may be misread in the data read by the optical character reading unit F102, and is misread. If there is a word including a possible character, the word registered in the first replacement table is read and replaced with the registered correct word to generate character data.

  The fourth step is the second replacement means F104, in which characters that may be erroneously read by the optical character reading means stored in the storage means are erroneously read from the registered second replacement table similar character list. Whether there is a character that may be misread for the first replacement character data other than the word data converted into the correct word obtained in the third step. The second replacement character data to be replaced with a mark that can be recognized as an arbitrary character that may be misread when it exists is generated.

  In the fifth step, the second replacement character data generated in the fourth step is generated as editing data by the editing data generation unit F105. The editing data generation unit F105 reads a dictionary table registered in the storage unit in advance, replaces a word including a mark that can be recognized as an arbitrary misread character with a correct word, and edit data Generate as

  The first replacement table F103 is a word including characters that are not registered in the first replacement table F103 and may be misread by the data storage unit F106, and the editing data generation unit F105 performs arbitrary misreading. The mark that can be recognized as a character that can be recognized is replaced with the correct word and the word before being replaced with the correct word are accumulated in the history table, and the accumulated update data is updated in the first replacement table F103. can do.

  The character data replacement method described above can be a recording medium on which a program that can be executed by a computer is recorded.

DESCRIPTION OF SYMBOLS 101 Data acquisition means 102 Optical character reading means 103 1st replacement means 104 2nd replacement means 105 Edit data generation means 106 Update data generation means 107 Replacement table update means 201 1st replacement table 202 2nd replacement table 203 History table 301 Image Data 302 Document data 303 Replaced data 304 Edited data 305 Update data 306 CSV data S101 Paper document digital conversion S102 OCR reading S103 Character existence check S104 Registered word?
S105 Automatic conversion to a correct word S106 Spell check S107 Conversion to text data for editing S108 Final character check S109 Data output for editing S201 Conversion to character recognition “〓” mark S202 Correction to a correct word S203 Replacement data stored in history table S204 Update Text data creation S205 Creation of update file S206 First replacement table data update S111 First replacement table S112 Second replacement table S113 Data storage history table S114 Similar word data addition / correction S115 Similar character list S116 Dictionary table

Claims (1)

In a method for replacing character data by a character data replacement system comprising a data storage unit and a data generation unit,
The data storage unit
A first replacement table, a second replacement table, and a history table;
The data generator is
Data acquisition means for acquiring image data in which characters are described;
Optical character reading means for reading character data acquired by the data acquisition means;
The data read by the optical character reading means is searched for whether or not there is a word containing a character that may be erroneously read, and if it exists, the word registered in the first replacement table First replacement means for replacing with,
A mark that can be recognized as a character that may be erroneously read by searching for a character that may not be replaced by the first replacement means and collating with a similar character list in the second replacement processing table. A second replacement means for replacing;
The data including the character replaced with the mark is replaced with a correct word, the replaced data is stored in the history table, and the character data is replaced with an editing data generating unit that generates the editing data. Method.

Images