JPH05342402A - Japanese character reader - Google Patents

Abstract

PURPOSE:To provide the Japanese character reader which can speedily and easily confirm whether the recognized result of the device is correct or not and can easily indicate correct answer characters without depending on Japanese syllabary (KANA)/Chinese character (KANJI) conversion or the like. CONSTITUTION:A correction part 3 to correct the recognized result of a recognition part 2 to read image data is provided with a function 3a to indicate character candidates having the high possibility of correction, function 3b to prepare correct answer characters or one part of correct answer characters by combining the constitutive elements of characters displayed by the device as the character candidates, function 3c to present relevant characters by retrieving a character pattern dictionary 5 based on the characters prepared by this function 3b, and function 3d to correct the recognized result into the character designated by an operator among the characters presented by this function 3c.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a Japanese character reader for optically reading Japanese characters.

[0002]

2. Description of the Related Art In recent years, Japanese characters have been used as a Japanese input means in a machine translation system for translating Japanese into other languages, and as a data input means when inputting document data in Japanese into a database or the like. Commercialization of a Japanese character reader for optically reading is actively performed.

In such a Japanese character reading device, various measures have been taken to increase the recognition rate. However, when the characters to be read include handwritten characters or dirty writing, etc. Has a recognition rate of 100%
Is technically very difficult to do.

Therefore, if the recognition result is erroneous, it is necessary to provide a correction function for correcting the error by a simple operation in order to improve the overall performance of the Japanese character reading device. It is an important point.

In the conventional case, as a method of correcting the recognition result, for example, all recognized character candidates are presented to the operator, and the operator corrects the recognition result by selecting with the mouse or the like. There is.

In addition, in case the correct character is not found in the character candidates selected by the apparatus, when the operator inputs a character code from the keyboard, a function for correcting the character corresponding to the character code is provided, or There has also been developed a system in which an operator inputs a phonetic reading of a character and the correct character is obtained by kana-kanji conversion.

[0007]

However, for example, when a foreigner who is unfamiliar with Japanese uses a Japanese character reading device, in the conventional device, whether the recognition result of the device is correct or not is The operator himself has to check the characters in the original text and the characters recognized by the device one by one, and it is necessary to check the characters one by one. Therefore, there is a problem that an enormous amount of time is spent and a great burden is imposed on the operator.

Further, when there is no correct answer character among the character candidates selected by the apparatus, the conventional method in which the operator checks a kanji table or the like and inputs the character code of the correct answer also imposes a burden on the operator. There is a problem that it is large and the workability is poor, and improvement has been desired.

Further, the method of allowing the operator to input a phonetic reading and obtaining the correct character by kana-kanji conversion is of little utility value if the operator is a foreigner who is not familiar with Japanese.

The present invention has been made in view of the above circumstances, and an operator can easily know a character that needs to be corrected without collating the character by character with the original, and whether the recognition result of the device is correct. Providing a Japanese character reading device that enables quick and easy confirmation of whether or not there is no correct character in the character candidates selected by the device, and it is easy without relying on Kana-Kanji conversion, etc. It is an object of the present invention to provide a Japanese character reading device capable of instructing the correct answer character to.

[0011]

According to a first aspect of the present invention, there is provided a Japanese character reader for optically reading characters in Japanese, and a reading unit for optically reading the characters and converting them into image data. A character pattern dictionary that stores information on character patterns and constituent elements of the character patterns, a word dictionary that stores information on constituent characters and parts of speech of words, and usage information, and image data sent from the reading unit to the character pattern dictionary. A character recognition unit that determines a character candidate for the read image data and an evaluation value for the character candidate by comparing each character pattern stored in The first candidate among the candidates is presented as a recognition result, and the correction process is performed on the recognition result of the recognition unit according to the instruction of the operator. Comprising a Tadashibu, and an output unit for outputting correction processing recognition results such as the reading result as a file or a higher-level data processing apparatus in which finished by the correcting unit.

Further, when presenting the first candidate obtained by the recognition unit as a recognition result to the correction unit,
An evaluation identification display function for clearly indicating that the evaluation value of the first candidate is equal to or less than the preset reference value and the character of which the difference between the evaluation values of the first candidate and the second candidate is equal to or less than the reference value is preset. Equipped with.

The Japanese character reader according to claim 2 is
A reading unit that optically reads Japanese characters, and that reads the characters optically and converts them into image data.
A character pattern dictionary that stores information on character patterns and constituent elements of the character patterns, a word dictionary that stores information on constituent characters and parts of speech of words, and usage information, and image data sent from the reading unit to the character pattern dictionary. A character recognition unit that determines a character candidate for the read image data and an evaluation value for the character candidate by comparing each character pattern stored in A first candidate among the candidates is presented as a recognition result, and a correction unit that executes a correction process on the recognition result of the recognition unit according to the instruction of the operator, and a recognition result after the correction process by the correction unit are performed. An output unit that outputs the read result to a file or a higher-level data processing device is provided.

Further, the correction section further comprises a character creation function, a correction candidate display function, and a candidate correction function.

Here, the character creation function, when modifying a character candidate, specifies the constituent elements of the character already displayed as the character candidate, combines the specified constituent elements, and modifies the character or a part of the character. create.

The correction candidate display function searches the character pattern dictionary based on the character or part of the character created by the character creating function, and presents all applicable characters.

The candidate correction function corrects the recognition result by the recognition unit to the character designated by the operator among the characters presented by the correction candidate display function.

[0018]

In the Japanese character reader according to claim 1,
When performing the correction process by the correction unit, character candidates that are likely to require correction in the recognition result by the recognition unit are
It is specified by the identification display by the evaluation identification display function.
Therefore, the operator can easily know the characters that need to be corrected without collating the recognition result with the original text one by one, and can quickly and easily confirm whether or not the recognition result of the device is correct. Therefore, the burden on the operator can be reduced.

Further, in the Japanese character reader according to the second aspect, when there is no correct answer character among the character candidates selected by the recognition unit, the character creation function and correction candidate display provided in the correction unit are displayed. By linking the functions such as the function and the candidate correction function, the operator can easily obtain the correct answer character only by operating the mouse on the screen of the display device, and the operator can specify the correct answer character by kanji characters. There is no need to look up the code table, and there is no need to rely on kana-kanji conversion, which forces the operator to enter phonetic readings.

Therefore, even if a foreigner who is unfamiliar with the Japanese language uses the Japanese character reader, the operator is not burdened, and the processing efficiency in the data conversion and translation processing of the Japanese document is improved. You can expect a significant improvement.

[0021]

1 is a block diagram of an embodiment of a Japanese character reading device according to the present invention. The Japanese character reader of this embodiment is used as a Japanese input means in a machine translation system for translating Japanese into another language, or as a data input means for inputting Japanese document data into a database or the like. As shown in the drawing, the reading unit 1, the recognition unit 2, the correction unit 3, the output unit 4, the character pattern dictionary 5, the word dictionary 6 and the like are used as constituent elements.

The reading unit 1 optically reads characters printed or handwritten on paper, converts them into image data, and sends them to the recognition unit 2. In addition, the character pattern dictionary 5 stores information such as character patterns, constituent elements of the character patterns, and arrangement positions of the constituent elements.
The word dictionary 6 stores information such as constituent characters of words, parts of speech, and usages.

The recognizing unit 2 compares the image data sent from the reading unit 1 with each character pattern stored in the character pattern dictionary 5 and the word stored in the word dictionary 6 and reads the image data. A character candidate for the image data and an evaluation value for the character candidate are obtained, and the obtained result is sent to the correction unit 3.

As will be described later in detail, the correction section 3 presents the first candidate among the character candidates sent from the recognition section 2 as a recognition result, and recognizes the recognition section 2 according to the operator's instruction. The correction process is executed on the result and the correction result is sent to the output unit 4.

The output unit 4 outputs the information such as the character string sent from the correction unit 3 as a reading result to a file or an upper data processing device.

The Japanese character reader of one embodiment optically reads Japanese characters written on paper or the like by the above components. FIG. 2 is a flow chart showing an outline of the reading process in the Japanese character reading device of the above embodiment.

Next, the flow of processing from reading the characters on the paper piece 7 shown in FIG. 3, correcting the reading result, and outputting to a file will be described with reference to FIG.

First, the reading unit 1 reads a character written on the paper piece 7 and sends it as image data to the recognition unit 2 (step 10). Then, the recognition unit 2 first performs image division processing (step 20). This image division processing divides the image data sent from the reading unit 1 into a group of characters such as paragraphs and lines, further divides the divided group of characters, and finally, It is divided into letters. FIG. 4 shows image data (character data) for each character obtained by the image division processing in step 20.

Next, the image data for one character divided in step 20 is matched with the character pattern stored in the character pattern dictionary 5 to calculate the degree of similarity (step 21). Here, the degree of similarity is 0 to 10
It is represented by an integer value of 0, and the larger the value, the better the similarity.

Then, a character having a similarity of 70 or more is set as a character candidate of the read character (image data), and the similarity is set as an evaluation value (character evaluation value) for the character candidate (step 22). Next, it is determined whether the image data processed in step 22 is the image data for the last one character divided in step 20, and if it is not the last image data, the image data to be processed Is updated to the next one character and the process returns to step 21 (steps 23 and 24), so that step 2
Steps 21 and 22 are repeated for all the image data divided by 0 for each character.

FIG. 5 shows the character candidates obtained by the processing in steps 21 to 24 and the evaluation value for each character candidate. In FIG. 5, the numbers 1, 2, ..., 9, ...
It corresponds to the order of image data for each character divided by
Character candidates for each image data are arranged in the vertical direction in descending order of evaluation value for each image data. The numerical value attached below the character shown as the character candidate is the evaluation value (similarity).

In the specific example shown in FIG. 5, "white" is selected as a character candidate for the first image data (image data for the first character "day" in FIG. 4).
Examples include "day", "month", "day", and the evaluation values for the character candidates are 92, 91, 78, and 72, respectively.

When the character candidates and the evaluation values thereof are obtained for all the image data (characters) by the steps 21 to 24, next, the word dictionary 6 for the words formed by combining the character candidates of the continuous image data. When a corresponding word is found in the word dictionary 6, the word is set as a word candidate, and the sum of the evaluation values of the character candidates forming the word candidate is calculated as the word evaluation value. (Step 25). For example, “white”, “day”, ... “Shiramoto”, “Shiraki”, “Japan”, which are word candidates or combinations of character candidates,
As a result of searching the word dictionary 6 for ..., “Japanese”, ...
When “white”, “day”, “month”, “shiraki”, “Japan”, “Japanese”, ... Are confirmed, those confirmed words are set as word candidates. Then, for example, for the word candidate “Shiraki”, the evaluation value 92 of the character candidate “white”
And 172, which is the sum of the evaluation value 80 of “tree”, are calculated as the word evaluation value.

In step 25, the process of searching the word dictionary 6 for words based on combinations of character candidates of continuous image data is performed for all combinations of the character candidates obtained in steps 21 to 24. FIG. 6 shows the word candidates obtained in the process of step 25 and the word evaluation value for each word candidate. In FIG. 6, the numbers 1, 2, ..., 9, ... Described in the uppermost row H correspond to the order of the image data for each character divided in step 20. The numerical value described below the character of each word candidate is the word evaluation value. Next, the possibility of connection between the word candidates obtained in step 25 is checked, and those that cannot be connected are deleted from the word candidates (step 26). It should be noted that words composed of one Kanji character are not consecutive.
Then, the first candidate, the second candidate, ...
And the order of character candidates is determined (step 27). The processing of steps 20 to 27 described above is performed by the recognition unit 2.

The correction unit 3 displays the guidance for correction processing and the image data read in step 10 on the screen of the display device, and the character string of the candidate character set as the first candidate in step 27. Is displayed on the screen of the display device as the recognition result, and further, for the designated character on the character string displayed as the recognition result, all candidate characters after the second candidate of the corresponding image data (character) are displayed. It is displayed on the screen of the device and the correction process is performed based on the instruction of the operator (step 30). The correction processing by the correction unit 3 will be described in detail later.

Next, the output unit 4 sets the character string of the first candidate after the correction processing by the correction unit 3 as a recognition result,
The data is output to a file or a higher-level data processing device according to the operator's instruction (step 40).

Above, a series of processes for reading Japanese characters is completed. Next, the functions provided in the correction unit 3 and the processing in the correction unit 3 will be described in detail. FIG. 7 shows a screen configuration displayed on the screen of the display device by the correction unit 3.

In FIG. 7, reference numerals 100 to 107 are guidance for instructing processing, 110 is an image display unit for displaying the image read by the reading unit 1 as it is, and 111 is an area in the image display unit 110. Area designation frame for display 120, a recognition result display unit for displaying the recognition result (first candidate) of the recognition unit 2 for the image data in the area specified by the area designation frame 111, and 121 the recognition result display A cursor for designating one character in the section 120, 130 is the cursor 1
It is a candidate display unit for displaying all the character candidates of the character designated by 21. 131 is the candidate display unit 1
Cursor for designating one candidate in 30; 140
Is enlarged image data near the character designated by the cursor 121.

The correction unit 3 has a function for correcting the candidate characters obtained by the recognition unit 2 on the correction screen of FIG. 7 as an evaluation identification display function 3a, a character creation function 3b, a correction candidate display function 3c, and a candidate. It is equipped with various functions such as the correction function 3d.

Here, when the evaluation identification display function 3a displays the character string by the first candidate as the recognition result of the recognition unit 2 on the recognition result display unit 120, the character candidate obtained by the recognition unit 2 is displayed. This is a function for clearly indicating that the evaluation value of the first candidate is equal to or less than a preset reference value and the character of which the difference between the evaluation values of the first candidate and the second candidate is equal to or less than the preset reference value.

In the case of this embodiment, the criterion for determining whether or not the evaluation value of the first candidate of the character candidates obtained by the recognition unit 2 is less than or equal to a preset criterion is the candidate character of the first candidate. If there is no word evaluation value, and if the word evaluation value does not exist, it is displayed so as to be distinguishable from a candidate having a word evaluation value by applying a dark shading. Note that the word evaluation value does not exist means that the candidate character does not exist in the word dictionary 6, and the probability that the candidate character is not the correct character is extremely high, that is, it should be corrected. It means that there is high quality.

Further, in the case of this embodiment, the criterion for judging whether or not the difference between the evaluation values of the first candidate and the second candidate is less than or equal to a preset criterion is the first candidate and the second candidate. The difference between the word evaluation values is 70 or less. If the difference is 70 or less, by applying a light shading, any of the candidates with a difference of 70 or more and the candidates with no word evaluation value exist. Also displayed in a distinguishable manner. In addition, when the difference between the word evaluation values of the first candidate and the second candidate is 70 or less, it means that there are a plurality of word candidates that are close to each other and have a high probability of being the correct character. It means that there is a high probability that the letters in question are not the correct letters.

Further, when the character creation function 3b corrects the character candidates, if the constituent elements of the character already displayed as the character candidate are designated, the designated constituent elements are combined and the character or a part of the character is designated. It is a function to create.

The correction candidate display function 3c is a function for searching the character pattern dictionary 5 based on the character or a part of the character created by the character creating function 3b and presenting all the applicable characters.

Then, the candidate correction function 3d has a function of displaying the characters displayed on the screen by the correction candidate display function 3c.
This is a function of correcting the recognition result by the recognition unit 2 to the character designated by the operator.

The flow of the correction process in the correction unit 3 utilizing the above-mentioned functions 3a to 3d will be described below with reference to FIG. The user determines whether or not the character that needs correction is in the current area specification frame 111. If the character that needs correction is not in the current area specification frame 111, the user sets the area specification frame 111. It moves (steps 301 and 302). Then, by using the mouse or the like, the cursor 121 is moved onto the character that needs to be corrected to select the corrected character (step 303). The correction unit 3 displays the character candidates of the character selected by the cursor 121 on the candidate display unit 130.

For example, when the operator selects "sentence" with the cursor 121, the correction unit 3 causes the candidate display unit 130 to display.
"Sentence", "Woman", and "Large" are displayed as the character candidates.
The operator looks at the character candidates displayed on the candidate display unit 130 and determines whether or not the correct answer character is included (step 304), and the correct answer character is displayed on the character candidate displayed on the candidate display unit 130. If it is included, the candidate to be changed is designated by moving the cursor 131 to the correct character (step 305). When the changed character is designated, the correction unit 3 corrects the recognition result to the character candidate designated by the cursor 131 (step 306).

On the other hand, when it is determined in step 304 that there is no correct character in the candidates displayed on the candidate display unit 130, the operator inputs the guidance character input process (inpu).
t character) is designated (step 307). When the character input process is designated by the operator, the correction unit 3 opens a character input window 350 as shown in FIG. 9, and the characters displayed in the candidate display unit 130 are displayed in the window 350. Display the candidates by dividing them into components.

In FIG. 9, the cursor 121 indicates that the character candidate is "magazine".
Text input processing of guidance 105 (input
character) is specified. In the window 350, the character “magazine” is divided into three constituent elements: “word” in the left half, “shi” in the upper right half, and “heart” in the lower right half.

Here, the operator designates necessary constituent elements from the constituent elements of the character displayed in the window 350 by using a mouse or the like, and the character input area indicated by a dotted line frame in the window 350. It moves to 351 and the correct answer character or a part of the correct answer character is created in the character input area 351 (step 308). The character creation processing in step 308 is performed by the character creation function 3b.

After completing the operation in step 308, the operator designates a character search. When the character search is designated, the correction unit 3 activates the correction candidate display function 3c. In the correction candidate display function 3c, the character pattern dictionary 5 is searched by using the character components created in the character input area 351 as keys, and all the corresponding characters are presented (step 309).

For example, as shown in FIG.
When “word” in the left half of “magazine” is created in the character input area 351 as a component of the correct answer character and a character search is designated, the correction candidate display function 3c uses the “word” as a key in the character pattern dictionary. 5 is searched, "total", "correction",
"Certificate", ..., "Accept", ... are displayed as the corresponding characters on the candidate display unit 130 and the like. In addition, as shown in FIG. 11, a character in which the left half of “magazine” and the right half of “sword” “nin” are combined as a constituent element of the correct answer character is created in the character input area 351. When the character search is designated, the correction candidate display function 3c searches the character pattern dictionary 5 by using "word" and "nin" as keys, and displays "recognition" as the corresponding character.

The operator uses the mouse or the like to specify the correct answer character among the characters displayed by the correction candidate display function 3c. Then, the above-mentioned candidate correction function 3d operates,
The recognition result is corrected to the character designated by the operator (step 310).

The operator corrects any character in the characters read by the reading unit 1 by moving the area designating frame 111 in the image display unit 110 and moving the cursor 121 in the recognition result display unit 120. If the recognition result to be corrected remains, the correction target is switched by the operation of moving the area designating frame 111 or the cursor 121, and steps 301 to 301 are executed.
When the process of 310 is repeated and the correction process for the recognition result is completed, if the end of the guidance (quit) is specified, the series of correction processes by the correction unit 3 ends (step 312).

In the Japanese character reading apparatus of the embodiment described in detail above, when the correction processing is performed by the correction unit 3, the character candidates that are highly likely to require correction within the recognition result by the recognition unit 2. Is clearly indicated by the identification display by the evaluation identification display function 3a. Therefore, the operator can easily know the characters that need to be corrected without collating the recognition result with the original text one by one, and can quickly and easily confirm whether or not the recognition result of the device is correct. It is also possible to reduce the burden on the operator.

When there is no correct answer character among the character candidates selected by the recognition unit 2, the character creation function 3b, the correction candidate display function 3c, and the candidate correction function 3 provided in the correction unit 3 are included.
By linking functions such as d, the operator can easily obtain the correct character by simply operating the mouse on the screen of the display device, and the operator checks the kanji code table to specify the correct character. There is no need, and there is no need to rely on kana-kanji conversion, which forces the operator to input phonetic readings.

Therefore, even when a foreigner who is unfamiliar with the Japanese language uses the Japanese character reader, the operator is not burdened and the processing efficiency in the data conversion and translation processing of the Japanese document is improved. You can expect a significant improvement.

The evaluation identification display function 3 of the correction section 3
In a, the criterion for determining whether or not the evaluation value of the first candidate of the character candidates obtained by the recognition unit 2 is less than or equal to a preset standard, and the difference between the evaluation values of the first candidate and the second candidate is determined in advance. The criteria for determining whether or not the criteria are less than or equal to the set criteria are not limited to those in the example. For example, the recognition unit 2
It is also possible to set an evaluation value of a character candidate or a specific numerical value of a word evaluation value of a word candidate as a criterion for determining whether or not the evaluation value of the first candidate of the character candidates obtained in step 1 is less than or equal to a preset standard. Conceivable. Further, as a criterion for determining whether or not the difference between the evaluation values of the first candidate and the second candidate is less than or equal to a preset standard, it may be possible to adopt a specific numerical value of the evaluation value of the character candidate. Further, it is also possible to set a value larger than the value (70) of the embodiment or a small value as a specific numerical value as a judgment criterion.

[0059]

As is apparent from the above description, in the Japanese character reading device according to the first aspect, when the correction processing is performed by the correction unit, it is necessary to correct the recognition result of the recognition unit. Character candidates with high property are clearly indicated by the identification display by the evaluation identification display function. Therefore, the operator can easily know the characters that need to be corrected without collating the recognition result with the original text one by one, and can quickly and easily confirm whether or not the recognition result of the device is correct. Therefore, the burden on the operator can be reduced.

Further, in the Japanese character reader according to the second aspect, when there is no correct answer character among the character candidates selected by the recognition unit, the character creation function and correction candidate display provided in the correction unit are displayed. By linking the functions such as the function and the candidate correction function, the operator can easily obtain the correct answer character only by operating the mouse on the screen of the display device, and the operator can specify the correct answer character by kanji characters. There is no need to look up the code table, and there is no need to rely on kana-kanji conversion, which forces the operator to enter phonetic readings.

Therefore, even when a foreigner unfamiliar with the Japanese language uses the Japanese character reading device, the operator is not burdened and the processing efficiency in the data conversion and translation processing of the Japanese document is improved. You can expect a significant improvement.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention.

FIG. 2 is a flowchart showing an outline of a reading process according to an embodiment of the present invention.

FIG. 3 is an explanatory diagram of an original text read in an embodiment of the present invention.

FIG. 4 is an explanatory diagram of image data subjected to image division processing according to an embodiment of the present invention.

FIG. 5 is an explanatory diagram of an example of selecting character candidates according to an embodiment of the present invention.

FIG. 6 is a calculation example of word candidates and word evaluation values according to an embodiment of the present invention.

FIG. 7 is an explanatory diagram of a screen display example by a correction unit according to an embodiment of the present invention.

FIG. 8 is a flowchart of a correction process in a correction unit according to an embodiment of the present invention.

FIG. 9 is an explanatory diagram of a screen display example by the correction unit according to the embodiment of the present invention.

FIG. 10 is an explanatory diagram of a screen display example by the correction unit according to the embodiment of the present invention.

FIG. 11 is an explanatory diagram of a screen display example by the correction unit according to the embodiment of the present invention.

[Explanation of symbols]

 1 reading unit 2 recognition unit 3 correction unit 3a evaluation identification display function 3b character creation function 3c correction candidate display function 3d candidate correction function 4 output unit 5 character pattern dictionary 6 word dictionary 7 paper piece 100 to 107 guidance 110 image display unit 111 area designation Frame 120 Recognition result display portion 121 Cursor 130 Candidate display portion 131 Cursor 140 Image data 350 Window 351 Character input area

Claims (2)

[Claims] 1. A Japanese character reader for optically reading characters in Japanese, comprising: a reading unit for optically reading characters and converting them into image data; and a character pattern and information on constituent elements of the character pattern. The stored character pattern dictionary, the word dictionary storing the constituent characters of the word, the part of speech, and the usage information, and the character patterns and the word dictionary in which the image data sent from the reading unit is stored in the character pattern dictionary A recognition unit that obtains a character candidate for the read image data and an evaluation value for the character candidate in comparison with the word stored in, and a first candidate among the character candidates obtained by the recognition unit as a recognition result. A correction unit that presents the correction result to the recognition result of the recognition unit according to the instruction of the operator, and a correction by the correction unit An output unit that outputs the processed recognition result as a reading result to a file or a higher-order data processing device, and the correction unit when presenting the first candidate obtained by the recognition unit as a recognition result. , An evaluation identification that clearly indicates that the first candidate has an evaluation value that is equal to or less than a preset reference value and the character that has a difference in evaluation value between the first candidate and the second candidate that is equal to or less than a preset reference value. A Japanese character reader characterized by having functions. 2. A Japanese character reading device for optically reading characters in Japanese, comprising: a reading unit that optically reads characters and converts them into image data; and a character pattern and information on the constituent elements of the character pattern. The stored character pattern dictionary, the word dictionary storing the constituent characters of the word, the part of speech, and the usage information, and the character patterns and the word dictionary in which the image data sent from the reading unit is stored in the character pattern dictionary A recognition unit that obtains a character candidate for the read image data and an evaluation value for the character candidate in comparison with the word stored in, and a first candidate among the character candidates obtained by the recognition unit as a recognition result. A correction unit that presents the correction result to the recognition result of the recognition unit according to the instruction of the operator, and a correction by the correction unit An output unit that outputs the processed recognition result as a reading result to a file or a higher-level data processing device, and the correction unit includes a character already displayed as a character candidate when correcting the character candidate. When a component of is specified, the specified component is combined to create a character or a part of a character, and the character pattern dictionary based on the character or part of the character created by the character creating function. And a correction candidate display function that presents all applicable characters, and a candidate correction function that corrects the recognition result by the recognition unit to the character specified by the operator among the characters presented by this correction candidate display function. A Japanese character reading device characterized by having and.