JP3452774B2 - Character recognition method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is a character string that is normally written on various kinds of slips, and is written at irregular character intervals or irregular writing methods, and contact and separation occur between adjacent characters. It relates to a technique for recognizing a low-quality character string that sometimes occurs.

[0002]

2. Description of the Related Art OCR for reading image data and converting it into character code data
(Optical character reading device) has been used in various businesses as its application field spreads. Different forms are used for each work, and the character strings and the people who fill in the forms are becoming various.

In the conventional OCR form, a character frame in which a character entry frame is printed character by character is used, and a large character frame is used particularly when Kanji is entered.
This is to make it easier for the OCR device to detect the characters to be entered one by one, and to encourage the writer to prevent the characters to be entered at the time of writing from contacting adjacent characters.

In the case of such a form, for example, even if the address and the name are entered in a few places, the number of characters to be entered becomes dozens of characters, and as a result, a large-sized form is required and the cost is high. In addition, it has been troublesome for the writer to fill in each frame one by one.

As the field of application of OCR spreads, a kanji character string can be entered in a small form such as an ordinary form without being restricted by a character frame, and can be recognized with practical recognition accuracy, and unreadable characters can be recognized. There is a need for character recognition / correction technology that enables efficient correction even when making corrections.

[0006] In the conventional typical character recognition method, the entered characters are referred to while referring to a file called a definition body that stores the coordinate position on the form of the character frame in which the character string to be recognized is entered. Each character is detected and cut out. Then, the recognition process is performed on each of the cut-out characters to output a candidate character group as a recognition result.

The recognition processing of the cut out characters is executed as follows, for example. First, a large number of unspecified writers collect the characters that have been filled in according to a predetermined format, the features that depend on the recognition method are extracted from these characters, and a statistical method (for example, clustering method) is used. A standard pattern is created. Then, a standard pattern dictionary is created from the target standard pattern for each character type.

The standard pattern is created as an average pattern obtained by averaging the collected character patterns, for example. More specifically, this average pattern is represented by the average feature amount obtained by calculating the average of the feature amounts corresponding to the collected characters.

In the process of recognizing handwritten characters, a large number of standard patterns are created for each character type because a large character shape change occurs depending on the writer. Usually, one standard pattern is called a template, and a dictionary created from a plurality of standard patterns for each character type is called a plural template dictionary.

The character recognition processing is executed using the above-mentioned standard pattern dictionary or plural template dictionary. Specifically, a feature amount is extracted from one character cut out from the input form, and between this feature amount and the feature amount of each template (standard pattern) that constitutes the standard pattern dictionary (or multiple template dictionary), Similarity or distance (Euclidean distance, Mahalanobis distance, etc.) is calculated. Then, each character type category to which each template up to a predetermined rank (e.g., 8th) belongs to a candidate character group in order of increasing similarity or decreasing distance.

When the recognized character is a character representing an address or a name, generally, knowledge processing using the address word and the name word is executed on the above candidate character group. More specifically, first, the candidate character strings for each writing position are combined in the entire writing position to output the candidate character string group.

Next, for each candidate character string constituting this candidate character string group, it is compared whether or not each word character string in the address dictionary or name dictionary of the knowledge processing target exists in the candidate character string. To be done.

Then, a score is given to the candidate character string according to the comparison result and the rank of each candidate character forming the candidate character string, for example. After this processing is executed for all the candidate character strings, the candidate character string with the highest score is output as the knowledge processing result.

As a conventional technique relating to such knowledge processing, for example, Japanese Patent Laid-Open Publication No. Sho 61-107.
The one disclosed in No. 486 is known. Here, when an address written in free pitch, a kanji character string such as a name, is recognized as in the case of writing on a regular slip, it is common that adjacent characters often come into contact with each other, and On the other hand, there are some characters that are written separately, such as Hen and Tsuku.

For this reason, it is difficult to judge which range is a range of one character by the conventional character recognition method in which the entered characters are detected, cut out and recognized one by one, and the recognition is practical. It is difficult to achieve accuracy.

Further, if each character cannot be recognized correctly,
In some cases, it may not be possible to determine how many characters have been entered,
In the conventional knowledge processing, which is premised on that the number of characters forming a word is fixed, there is a limit in improving the recognition accuracy.

Further, particularly in the recognition processing of an address place name, when a higher-level word (for example, Tokyo, Osaka prefecture, etc.) cannot be recognized by the knowledge processing, the lower-level word is processed by the knowledge processing at that stage. Since it is generally not done, to correct the address place name, 1
It is necessary to correct all the character strings starting from the first character.

As a first conventional technique for recognizing the above-mentioned free-pitch character string, the one disclosed in Japanese Patent Publication: Japanese Patent Publication No. 8-23875 "word reading method" is known. There is. In this first conventional technique,
The candidate character string which is the recognition result and the word dictionary are collated by DP matching or the like, a word having a large number of matching characters is selected, the mismatched portion is cut out again, and the cut-out character string is further recognized. Be seen.

As a second conventional technique for recognizing a free-pitch character string, Japanese Patent Laid-Open Publication No. 63-63.
The one disclosed in No. -136291 "Word reading method" is known. In the second conventional technique, the recognition processing is executed by using a standard pattern dictionary having partial patterns indicating the respective portions of the character deviation and the character string as a standard pattern, and the character string from the character deviation and the character string of the candidate character string is calculated. Is generated, and the matching process with the word dictionary is executed.

As a third conventional technique for recognizing a free-pitch character string, Japanese Patent Laid-Open Publication No. 8-
The one disclosed in No. 171614 “Character string reading device” is known. In the third conventional technique, a predicted character is detected when a skipped character occurs because the correct character is not included in the candidate character string, or when a plurality of read candidates occur due to the existence of a character candidate that conflicts with the correct character. The existence of the column is verified. As a verification means in this case, some implementation methods are disclosed.

However, the recognition processing for a character string that we normally write, that is, a contact between adjacent characters frequently occurs, the character width changes greatly for each character, and a low-quality character string that is often crushed or faint is recognized. The above-mentioned first to third prior arts have the following problems.

First, in the first conventional technique, it is uncertain which character of the candidate character string is preferentially treated, and all the characters in the candidate character string are treated equally. There is a problem in that there is a possibility that only inappropriate words may be selected as candidates.

Next, in the second conventional technique, there is a problem in processing the area where adjacent characters are in contact with each other. Furthermore, the third
In the related art, several methods are described as a method of realizing the verification means, but all of them use a combination of character candidates, and their verification performance greatly depends on the cutting result of the first character. It has a problem of dependence.

An object of the present invention is to accurately recognize a low-quality character string by paying attention to a specific character.

[0025]

SUMMARY OF THE INVENTION The present invention provides a character recognition / correction method for recognizing characters constituting an input character string entered in an entry field having a predetermined category, or a computer-readable record having a function equivalent thereto. Assuming medium.

In the present invention, first, the input character string and the first
By executing the first matching process with the recognition dictionary (specific character standard pattern dictionary 107), the specific character or the specific character string is extracted from the input character string. More specifically, a standard pattern corresponding to a specific character or a specific character string is stored in the first recognition dictionary, and the first pattern is stored between the pattern of the input character string and each standard pattern in the first recognition dictionary. By executing the matching process of, the specific character or the specific character string is extracted from the input character string. The above-mentioned specific character or specific character string is, for example, one that frequently appears in a predetermined category or one that has high recognition accuracy.

Next, a candidate that belongs to a predetermined category (for example, an address character string) and is possibly located in an area in the input character string before or after each specific character extracted from the input character string. A word group is extracted from the category-based word dictionary (specific character dictionary 110, knowledge dictionary 111).

Then, for each candidate word belonging to the extracted candidate word group, the second recognition is performed for each area in the input character string in which each candidate word is located based on the information about each candidate word. By performing the second matching process using the dictionary (standard pattern dictionary 113), the characters that form the input character string are recognized. More specifically, the second recognition dictionary stores standard patterns corresponding to characters or character strings related to the candidate words belonging to the candidate word group, and for each candidate word belonging to the candidate word group, each candidate A second matching process is performed between the pattern of each candidate word and each standard pattern in the second recognition dictionary for each area in the input character string where each candidate word is located based on the information about the word. By being executed, the characters forming the input character string are recognized. In this case, for example, information on the number of characters of each candidate word is used as the information on each candidate word. Also, the second recognition dictionary may be configured to include the first recognition dictionary.

According to the configuration of the above-described invention, the specific character or the specific character string in the input character string is first preferentially recognized, the candidate words before and after it are assumed based on the recognition result, and the information of the candidate word is further assumed. Characters that make up the input character string entered at irregular intervals and entry methods that are normally found on various forms (slips) by re-recognizing the characters that make up the input character string using Can be recognized with high accuracy.

In the above configuration of the invention, the recognition result of the characters forming the input character string is displayed in parallel with the input character string, and the desired area on the displayed input character string is designated by the user and the desired area is displayed. The character or character string corresponding to the area is corrected, and based on the information about the correct character or correct character string given by the correction, the candidate word group extraction process and the second matching process are executed again, and the input character string It can be configured so that the characters that make up the are recognized again. In this case, in response to designation of the desired area on the displayed input character string, a plurality of candidate recognition results in the desired area can be displayed.

With such a character correction technique, other unrecognizable parts can be automatically corrected only by correcting a specific character or character string. Further, in the above-described configuration of the invention, a word having a notational fluctuation for each candidate word may be output as a new candidate word belonging to the candidate word group.

With such a notational fluctuation control technique, various entry methods can be flexibly dealt with.

[0033]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail below with reference to the drawings. Configuration and General Operation of Embodiment of the Present Invention FIG. 1 is a configuration diagram of an embodiment of the present invention.

First, the character slicing unit 103 uses the entry field definition 104 defining the information on the entry field position of the form, from the beginning of the input character string 101 entered in the form read from the image memory 102. Cut out one character at a time.

Next, the feature extraction unit 105 extracts a feature amount from the cut out character. Subsequently, the matching unit 106 performs a matching process between the feature amount of the cut out character and the feature amount of each specific character standard pattern in the specific character standard pattern dictionary 107, and the matching degree is determined in descending order of matching degree. The character string category of each specific character to which each specific character standard pattern up to the rank belongs is set as a candidate specific character for the extracted character as a candidate character string buffer 1
It outputs to 08.

Character cutout unit 103, feature extraction unit 10
5 and the series of specific character recognition processing by the matching unit 106, the character cutting unit 103 performs the input character string 101.
It is executed for each character cut out in order from the beginning. As a result, the input character string 101 is stored in the candidate character string buffer 108.
The candidate specific characters corresponding to the respective characters are held in the order corresponding to the order of the characters cut out from.

The candidate word search unit 109 extracts all sets (specific character sets) consisting of any two adjacent specific characters from the candidate specific character strings obtained in the candidate character string buffer 108, and specifies each of them. It is searched whether the character set is registered in the specific character dictionary 110.

When one specific character set is registered in the specific character dictionary 110, the candidate word searching unit 109 forms the specific character set from the record in the knowledge dictionary 111 linked to the registered record. A word group sandwiched by two specific characters is searched, and the searched word group is held in the candidate word buffer 112 as a candidate word group.

The candidate word search unit 109 extracts a candidate word group corresponding to each of the above-mentioned specific character sets extracted from the candidate character string buffer 108, and holds it in the candidate word buffer 112.

After all, in the candidate word buffer 112, one or more candidate word groups are obtained for one specific character set,
Eventually, a set of candidate word groups for a plurality of specific character sets will be obtained.

Each candidate word belonging to the candidate word group obtained in the candidate word buffer 112 for one specific character set is
It is sequentially read and the following series of processing is executed for each.

First, the character slicing unit 103 constructs a specific character set to which the candidate word belongs in the input character string 101 read from the image memory 102, using the information of the candidate word output from the candidate word buffer 112. The character string in the character string area sandwiched between two specific characters is cut out again.

The feature extraction unit 105 extracts a feature amount from the re-cut character string. Furthermore, the matching unit 106
Performs matching processing of the feature amount of the re-cut out character string and the feature amount of each standard pattern in the standard pattern dictionary 113, which is the second dictionary. The category of the character string to which the standard pattern belongs is held in the candidate character string buffer 108 as a candidate recognition result group for the candidate word.

Character cutout unit 103, feature extraction unit 10
5 and the series of re-recognition processing by the matching unit 106 is executed for each candidate word belonging to the candidate word group obtained in the candidate word buffer 112 for the one set of specific character set, and predetermined for each candidate word. A candidate recognition result group up to the rank is obtained in the candidate character string buffer 108.

Then, the matching unit 106 selects the most appropriate and reliable among all the candidate recognition result groups up to a predetermined rank obtained in the candidate character string buffer 108 for each candidate word belonging to the one set of specific character set. The recognition result having a high degree of matching, more specifically, the candidate recognition result having the highest degree of matching is
The result is output to the knowledge processing unit 114 as the recognition result of the portion sandwiched between the two specific characters forming the specific character set of the set.

Character cutout unit 103, feature extraction unit 10
5 and the series of re-recognition processes by the matching unit 106 for each candidate word belonging to the candidate word group of one specific character set are executed for each specific character set registered in the candidate word buffer 112. . As a result, the knowledge processing unit 114 outputs the recognition result corresponding to each character area sandwiched between the two specific characters forming each specific character set.

The knowledge processing unit 114 performs knowledge processing using the entry field definition 104 and the knowledge dictionary 111 on the recognition result corresponding to each character region sandwiched between two specific characters forming each specific character set. , The final recognition result of the entire character area including the above-mentioned character areas is determined, and is output to the recognition result buffer 115.

In the series of recognition processes described above, reject (unrecognizable) information is added to the part of the character or character string that does not satisfy the recognition condition to the end. In this case, the recognition result obtained in the recognition result buffer 115 is displayed on the display unit 117 via the interface unit 116. When the user displays the recognition result on the display unit 117, the user uses the input unit 11 including a mouse and a keyboard.
From 8, it is possible to correct unrecognizable characters / character strings.

The user only has to specify a specific correct character in the unrecognizable character / character string from the input unit 118, and information about the correct character is output from the interface unit 116 to the correct character buffer 119 and the area coordinate buffer 120. It

The candidate word searching unit 109 uses the information about the correct character obtained in the correct character buffer 119 as the information of the specific character and the specific character dictionary 110 and the knowledge dictionary 1 described above.
By executing the candidate word search process using 11, the unrecognizable character can be correctly recognized again.
Further, the character cutout unit 103 determines the cutout position of the correct character specified by the user as the area coordinate buffer 120.
The correct character segmentation can be performed by acquiring from.

As described above, in the present embodiment, with respect to the input character string 101 such as the address, the name, the product name, etc. written in each entry field in the form, a character having a high frequency of appearance in each field or a specific appearance frequency is specified. By focusing on the character / character string,
Using the word information held in the knowledge dictionary 111 and the connection information for each character area in the case of a character string such as an address having a hierarchical structure, it is possible to select a candidate word in the character area sandwiched between the specific characters. it can. Further, in the present embodiment, the information of the candidate word is used to extract the character area sandwiched between the specific characters from the input character string 101 and to perform the re-recognition process on the character area, so that the adjacent character A character string written in a style that often causes contact and separation between
It can be recognized with high recognition accuracy. Detailed Operation of the Embodiment of the Present Invention FIGS. 2 to 4 are operation flowcharts showing overall control realized by the embodiment of the present invention having the configuration shown in FIG. < Specific Character Recognition Processing > First, the character cutout unit 103
Using the entry field definition 104 that defines the information about the entry field position of the form.
One character is cut out from the beginning of the input character string 101 written on the form, which is read out as binary image data from No. 02 (step 201 in FIG. 2).

FIG. 5 is a diagram showing an example of the data format of the entry field definition 104 used by the character slicing section 103. For example, when fields 1 and 2 are arranged on the form and the character strings entered in these two fields are recognized, the entry field definition 104 is determined as follows.

First, the upper part of the form is the coordinate origin, the x-axis is defined in the horizontal direction, and the y-axis is defined in the vertical direction. For each of fields 1 and 2, the coordinates of the upper left corner of the field (field origin) are defined. Coordinate size), and field size data consisting of a field width in the x-axis direction and a field height in the y-axis direction are defined as shown in FIG. 5 (a). The unit of length is millimeters or inches.

Next, for each of the fields 1 and 2, a field type that indicates what kind of character string is written in each field is defined. These pieces of information are held in the storage device (not shown) as the entry field definition 104 in the table format shown in FIG.

The character slicing unit 103 determines the character area for each field on the image data read from the image memory 102 by using the above-described entry field definition 104, and then the image data in the character area is determined. On the other hand, the character cutout control shown by the operation flowchart shown in FIG. 6 is executed.

Here, as shown in FIG. 8A, the field origin coordinates of the target area extracted from the entry field definition 104 are (x ₀ , y ₀ ), and the field width in the x-axis direction is dx, y. The field height in the axial direction is dy.

First, the character slicing section 103 accumulates the number of black pixels for each scanning line in the x-axis direction to obtain a horizontal histogram showing the appearance frequency of black pixels in the x-axis direction at each y-coordinate position. , As shown in FIG. 8B (step 601 in FIG. 6).

Next, as shown in FIG. 8 (b), the character slicing unit 103 scans the horizontal histogram from above and below, respectively, and first finds the positions α and β that exceed the frequency value C. The value α-β calculated and further calculated therefrom is set as the character string height h in the target area (step 602).

Next, the character slicing section 103 accumulates the number of black pixels for each scanning line in the y-axis direction,
A vertical histogram showing the appearance frequency of black pixels in the y-axis direction at each x-coordinate position is calculated as shown in FIG. 8C (step 603 in FIG. 6).

Subsequently, as shown in FIG. 8C, the character clipping unit 103 scans the vertical histogram from the left, and the frequency value changes from the threshold value d or less to the threshold value d or more. Points x ₁ , x ₃ , x ₅ , ... (x _2n-1 : n
= 1, 2, ...) is calculated as a cutout candidate position,
Further, points x ₂ , x ₄ , x ₆ , ... (x _2m : m = 1, where the frequency value changes from the threshold value d or more to the threshold value d or less)
2, ...) are also calculated as cutout candidate positions (step 604).

Next, the character slicing section 103 calculates a region [x _2m , x _2n-1 ] satisfying the following conditional expression, and sets it as a character slicing result (step 605).

[0062]

## EQU1 ## h−t ₁ ≦ x _2m −x _2n−1 ≦ h + t ₂ (m = 1, 2, 3, ...), (n = 1, 2, 3, ...
Here, h is the height of the character string calculated in step 602 described above, and t ₁ and t ₂ are parameters determined by the learning sample of the input character string 101. In the example of FIG. 8 (c), the following three areas are calculated as the character extraction result. [X ₁ , x ₂ ] [x ₃ , x ₄ ] [x ₅ , x ₈ ] As a result of the processing of step 605, the character segmenting unit 103 determines whether or not there is an area that satisfies the following conditional expression (step 606).

[0063]

[Number 2] _{x 2l -x 2l-1> h} + t 2 (l = 1,2,3, ···) If the determination in step 606 is NO, the character segmentation unit 1
03 ends the control process.

If the determination in step 606 is YES, the character clipping unit 103 determines that the frequency value of the vertical histogram calculated in step 603 is larger than the threshold value d in the region [x _2l-1 , x _2l ]. Below a specified value, and
A value k that satisfies the following conditional expression is calculated.

[0065]

## _EQU00003 ## h.apprxeq. ( _{X.sub.2l-x.sub.2l-1} ) / k As a result, each position obtained by dividing the region [ _x.sub.2l-1 , x _2l ] into k is calculated as the character cut-out position (above, step 60).
7). In the example of FIG. 8 (d), l = 1 and k = 2, and the position x ′ obtained by dividing the area [x ₁ , x ₂ ] into two is calculated as the character cut-out position.

After that, the character slicing section 103 ends the control processing. The operation flowchart of FIG. 6 described above corresponds to the character cutout process executed by the character cutout unit 103 for a field in which the number of characters is not given in advance.

On the other hand, as in the case where the re-recognition process is executed based on the information of the candidate words read from the candidate word buffer 112, the character slicing unit 103 causes the character slicing target area and its area In some cases, the character cutting process may be performed on a field in which the number of characters in the field is given in advance.

In this case, the character cutout unit 103
Instead of the processing group of steps 605 to 607 of FIG. 6, the processing of step 701 of FIG. 7 is executed. That is, the x coordinate at the left end of the area to be cut out is x _s , and the x coordinate at the right end is x.
When the coordinate is x _t and the number of characters in the area is n, the character segmentation unit 103 performs step 60 in FIG.
A position where the frequency value of the vertical histogram calculated in 3 is equal to or less than a predetermined value and has an adjacent interval close to a value X _n that satisfies the following conditional expression is calculated as a character cutout position.

[0069]

## EQU00004 ## (x _t −x _s ) / n = X _n Specifically, two adjacent character cutting positions are defined as x _i ,
x _{i + 1} (i = 1, 2, ..., X _s ≦ x _i , x _{i + 1} ≦ x
_t ), the character cutout unit 103 determines that the character cutout position x _i (x _i ≠ x _s , x _t ) that satisfies the following conditional expression.
To calculate.

[0070]

X _n −t ₅ ≦ x _{i + 1} −x _i ≦ X _n + t ₆ where t ₅ and t ₆ are parameters determined by the learning sample of the input character string 101.

After the character cutout processing by the character cutout unit 103 described above, the feature extraction unit 105 extracts a feature vector, which is a feature amount for recognition, from the cut-out one character (step 202 in FIG. 2). ).

Specifically, the feature extraction unit 105 extracts the feature vector by the following series of processing, for example. That is, first, the feature extraction unit 105 extracts a character contour pixel from the cut-out character image data.

Next, the feature extraction unit 105 divides the cut out area into a plurality of divided areas. Further, the feature extraction unit 105 extracts a direction component (for example, a four-direction component of a vertical direction, a horizontal direction, a diagonal left direction, and a diagonal right direction) for each contour pixel in the divided region for each divided region, By summing up the direction components of all the contour pixels in the divided area, a total value for each direction component in the divided area is calculated, and a partial feature vector having them as element values corresponding to each direction component is calculated.

Finally, the feature extraction unit 105 extracts the feature vector by integrating the elements of the partial feature vectors of all the divided areas. After the feature extraction unit 105 has extracted the feature vector of the extracted character as described above, the matching unit 106 extracts the feature vector of the extracted character and each specific character standard pattern in the specific character standard pattern dictionary 107. The matching process with the feature vector is performed (step 203 in FIG. 2), and the character type categories of the specific characters to which the specific character standard patterns up to a predetermined rank belong in the descending order of matching degree are extracted as the extracted characters. It is output to the candidate character string buffer 108 as a candidate specific character group for (step 204 in FIG. 2).

More specifically, the matching unit 106
For example, a distance (Euclidean distance, Mahalanobis distance, etc.) is calculated between the cut-out character feature vector and the feature vector of each specific character standard pattern in the specific character standard pattern dictionary 107. Then, the matching unit 106 determines the character type categories of the specific characters to which the specific character standard patterns up to a predetermined rank (nth) belong in the order of increasing distance as candidate character groups for the cut-out specific characters. Output to the column buffer 108.

If the distance of the standard pattern of the first specific character is larger than the predetermined threshold value T ₁ , rejected (unrecognizable) information is added to the cut out character. Here, the above-mentioned specific character standard pattern dictionary 107 will be described with a specific example.

Now, consider the case where the input character string 101 is an address character string. In the present embodiment, at first, for example, in an address character string, the frequency of appearance that indicates the division of the hierarchical structure is high, and "city", "road", "prefecture", "prefecture", "city", "ward", "county".
Only specific characters consisting of one or two characters such as “town”, “village”, “letter”, “large letter” need to be recognized. Further, in the address character string, specific characters such as “east”, “west”, “south” and “north” also appear frequently.

Therefore, in the present embodiment, in order to improve the recognition accuracy of these specific characters, the specific character standard pattern dictionary 107 which is composed of only the standard patterns of these specific characters and has a small dictionary capacity is used.

Such a specific character standard pattern dictionary 10
By preparing 7 separately from the standard pattern dictionary 113, the recognition processing speed can be shortened and the recognition accuracy can be improved.

The specific character standard pattern dictionary 107 is constructed as the same dictionary as the standard pattern dictionary 113, and many templates (standard patterns) are stored for each specific character in order to improve the recognition accuracy of the specific character. May be configured as.

On the other hand, when the input character string 101 is a name character string, there are no delimiter characters like the address character string, but there are character types that appear frequently. For example, the characters used as surnames are in the top 50 in terms of appearance frequency.
Since the character types up to the 0th place cover about 82%,
The specific character standard pattern dictionary 107 can be constructed with the upper N characters.

Alternatively, it may be constructed such that only N character types are selectively used from the standard pattern dictionary 113 for recognition of a specific character. Then, the specific character dictionary 110 is configured to correspond to the above-mentioned specific character type.

Further, instead of selecting the character type according to the appearance frequency, it is possible to statistically determine easily recognizable characters from a large number of actual data and select the determined character type. .

Character cutout unit 103, feature extraction unit 10
5 and the series of specific character recognition processing by the matching unit 106, the character cutting unit 103 performs the input character string 101.
It is executed for each character cut out in order from the beginning of the (repeating steps 205 → 202 in FIG. 2). As a result, the candidate character string buffer 108 holds a candidate specific character group for each character in an arrangement order corresponding to the arrangement order of the characters cut out from the input character string 101. < Search for candidate words in the specific inter-character area and re-recognize in that area
Knowledge processing > the candidate word search unit 109 extracts all sets of adjacent two specific characters (specific character sets) from the set of candidate specific character groups obtained in the candidate character string buffer 108, and extracts each of them. The specific character set of is a specific character dictionary 110
Search whether or not registered in. Then, the candidate word search unit 109 determines that one set of specific characters is the specific character dictionary 11
When registered in 0, the record in the knowledge dictionary 111 linked to the registered record is searched for a word group sandwiched by two specific characters forming the specific character set, and the searched word group is a candidate. The word group is held in the candidate word buffer 112 (above, step 20 in FIG. 3).
6).

Now, consider the case where the input character string 101 is an address character string. Note that name character strings, product name character strings, and the like other than the address character string do not have a hierarchical structure, and thus can be realized in the same manner as the case of the address character string except for the portion related to the hierarchical structure.

The structure of the knowledge dictionary 111, which is an address dictionary, is
For example, as shown in FIG. 10, according to the hierarchical structure of the address, the words are divided into level 1: prefecture, level 2: city / gun, level 3: town / village, ... It is stored.

On the other hand, in the specific character dictionary 110, as shown in FIG. 11, for each record corresponding to a specific character set consisting of two specific characters "character 1" and "character 2",
A data set including pointer information and data number information starting from the pointer for indicating a set of records in the knowledge dictionary 111 in which a word group sandwiched by two specific characters forming the specific character set is stored. It is stored. As this data set, a plurality of sets can be designated as shown in FIG. 11. In the record for each specific character set of the specific character dictionary 110, as shown in FIG. The pointer number information N corresponding to the number of data sets of information is also stored.

In the example of FIG. 12, a record corresponding to a specific character set composed of two specific characters of a blank character and “prefecture” in the specific character dictionary 110 has a level in the knowledge dictionary 111 shown in FIG. A data set (pointer information and data number information) indicating n ₁ records starting from the word “Aomori” in one area and n ₂ records starting from the word “Kanagawa” in the level 1 area, respectively, and a pointer. The number N = 2 is registered.

Further, in the example of FIG. 13, the specific character dictionary 110
In the record corresponding to the specific character set consisting of two specific characters, “to” and “ward”, in the record n ₃ starting from the word “Chiyoda” in the level 2 area in the knowledge dictionary 111 shown in FIG. The number of records and the number of pointers N = 1 are registered.

The address usually ends with the notation "... chome ... address ... kata", but with such special characters "chome""address""ban""kata""go" In many cases, numbers are entered in the sandwiched area instead of words. In such a case, as shown in FIG. 14, in the record corresponding to the specific character set made up of the specific characters in the specific character dictionary 110, the data including the pointer information and the data number information as described above. Not a pair, but "(number) *
A symbol such as "n" is set. Candidate word search unit 1
When searching for a record in which the above-described symbol is set from the specific character dictionary 110, it is confirmed that the numbers 09 are continuously entered in the area sandwiched between the specific characters as described above. The detection result indicating the detection is written in the candidate word buffer 112.

Furthermore, the specific character dictionary 110 and the knowledge dictionary 111 as shown in FIG. 15, for example, can be configured. That is, in the example of FIG. 15, in the specific character dictionary 110,
In a record corresponding to a specific character set consisting of two specific characters, that is, a blank character and “kawa”, pointer information indicating the 4-character word “Kanagawa” in the knowledge dictionary 111 and the number of data = 1, and the knowledge dictionary 111. The pointer information pointing to the two-letter word “Kanna” and the number of data = 1 are set.

The record corresponding to a specific character set consisting of two specific characters “kawa” and “middle” in the specific character dictionary 110 indicates the two-character word “Sakiichi” in the knowledge dictionary 111. The pointer information and the number of data = 1 are set.

Further, in the record corresponding to the specific character set composed of two specific characters “middle” and “middle” in the specific character dictionary 110, the five-character word “Hara-ku Ueda” in the knowledge dictionary 111 is recorded. Pointer information pointing to and the number of data = 1 are set.

As described above, it is possible to store the information corresponding to the specific characters and words that frequently appear in the address character string in the specific character dictionary 110 and the knowledge dictionary 111. Next, as shown in FIG. 16, the specific character dictionary 11
If the word "Marunouchi" in the knowledge dictionary 111 is linked to the record corresponding to the specific character set consisting of two specific characters, "ku" and the specific character indicating the end of the address, in 0, As a fluctuation, the character string "Marunouchi" may be entered instead of "Marunouchi". In such a case, it is useless to store the words corresponding to all the fluctuations of the writing in the knowledge dictionary 111.

Therefore, in the present embodiment, when a word in the knowledge dictionary 111 linked from the specific character dictionary 110 is searched, a control operation for coping with the fluctuation of the notation shown in the operation flowchart of FIG. 9 is executed. .

First, the candidate word searching unit 109, for one set of specific character sets, the specific character dictionary 110 and the knowledge dictionary 11
1 is searched according to the rules described so far, and the word group searched as a result is written in the candidate word buffer 112 as a candidate word group corresponding to the specific character set currently being processed (step 901 in FIG. 9). This step 901 is part of step 206 of FIG.

Next, as a part of step 206 in FIG. 3, the candidate word search unit 109 performs a drawing for each candidate word group obtained in the candidate word buffer 112 for one specific character set. The series of processes shown in Steps 902 to 910 of 9 are repeatedly executed.

That is, the candidate word search unit 109 changes the hiragana to katakana when the characters forming the detected word have hiragana, and the word obtained as a result is changed to the specific character set currently being processed. The corresponding candidate word is written in the candidate word buffer 112 (step 902 → 903 in FIG. 9).

Next, the candidate word search unit 109 changes the katakana to hiragana if the characters forming the detected word have katakana, and the word obtained as a result is changed to the specific character set currently being processed. Is written in the candidate word buffer 112 as another candidate word corresponding to (step 904 → 905 in FIG. 9).

Next, the candidate word searching unit 109 changes the Chinese numeral into an Arabic numeral when the detected word has a Chinese numeral, and the resultant word is currently being processed. Other candidate words corresponding to the specific character set are written in the candidate word buffer 112 (steps 906 → 907 in FIG. 9).

Next, the candidate word searching unit 109 determines that if the characters forming the detected word include Arabic numerals,
The Arabic numeral is changed to the Chinese numeral, and the word obtained as a result is written in the candidate word buffer 112 as another candidate word corresponding to the specific character set currently being processed (steps 908 → 909 in FIG. 9).

Finally, the candidate word searching unit 109 determines that the characters forming the detected word can be omitted characters (for example, "Mizonokuchi").
When there is an abbreviated "Mizoguchi"), the character string obtained by omitting the abbreviated characters is used as another candidate word corresponding to the specific character set currently being processed.
Write to the candidate word buffer 112 (step 9 in FIG. 9).
08 → 909).

If there is a candidate word group that has not yet been subjected to the control processing for the fluctuation of the notation in the candidate word buffer 112 for one set of specific character sets, the candidate word searching unit 109 of FIG. A series of processing shown in steps 902 to 910 is repeatedly executed (step 91 in FIG. 9).
1 → 902 to 910 → 911).

As described above, for the candidate word group obtained in the candidate word buffer 112 for one specific character set, the control for the fluctuation of the notation is realized. As described above, a candidate word group is obtained in the candidate word buffer 112 for one specific character set selected from the candidate character string buffer 108.

Now, for example, the input character string 1 shown in FIG.
When 01 is entered, steps 201 to 2 in FIG.
Through the specific character recognition processing of 05, the area 1701 is recognized as the specific character “Tou” and the area 1702 is recognized as the specific character “ward”.

For this recognition result, the candidate word search unit 1
In step 206 of FIG. 3 described above, 09 is a record of a specific character set consisting of a blank character and a specific character “TO” in the specific character dictionary 110, and from the entry in the knowledge dictionary 111 linked to the registered record. , Searching for one word “Tokyo” sandwiched by two specific characters forming the specific character set, and corresponding the searched word to a specific character set consisting of a blank character and a specific character “TO” The candidate word group 112 holds the candidate word group. In this case, the number of candidate word groups for the specific character set is one, and as shown in FIG. 18, the candidate word "Tokyo" has two characters.

In addition, the candidate word search unit 109 executes the second time in step 206 after the determination in step 211 of FIG. 3, which will be described later, in the specific character dictionary 110. A record of a specific character set consisting of is detected, and from the entry in the knowledge dictionary 111 shown in FIG. 10 linked to the registered record, 23 words “Chiyoda” sandwiched by two specific characters forming the specific character set are detected. , “Center”, “port” ... Are searched and the searched word group is held in the candidate word buffer 112 as a candidate word group corresponding to the specific character set. In this case, the number of candidate word groups for the specific character set is 2
There are three, and as shown in FIG. 19, the number of characters of each candidate word is either three, two, or one.

In this way, the candidate character string buffer 10
After a candidate word group is obtained in the candidate word buffer 112 for one specific character set selected from 8, the character segmentation unit 103, the feature extraction unit 105, and the candidate extraction unit 105 for each candidate word belonging to the candidate word group. And matching unit 106
However, a series of re-recognition processes in steps 207 to 211 of FIG. 3 is executed to extract a candidate recognition result group up to a predetermined rank for each candidate word.

First, the character slicing unit 103 constructs a specific character set to which the candidate word belongs by using the information of the candidate word output from the candidate word buffer 112 in the input character string 101 read from the image memory 102. The character string in the character string area sandwiched between two specific characters is cut out again (step 207 in FIG. 3).

In this case, if the number of characters of the candidate word is,
If there are two characters such as “Tokyo” shown in FIG. 8 or “center” shown in FIG.
In accordance with the operation flow chart shown in steps 601 to 604 of FIG. 6 and step 701 of FIG. Determine the cutout position of.

When the number of characters of the candidate word is three, such as "Chiyoda" shown in FIG. 19, the character cutout unit 103 sets the target area of the character cutout to three.
After dividing (n = 3 in the above-mentioned formula 3), the cut-out position of each character is determined.

Further, when the number of characters of the candidate word is one character like "port" shown in FIG. 19, the character slicing section 103 assumes that only one character is present in the area to be the character slicing target. (N in the above equation 3
= 1).

Next, the feature extraction unit 105 extracts a feature vector for each character of the re-cut character string as described above (step 208 in FIG. 3). Furthermore,
The matching unit 106 includes, for each character, the feature vector of the character and the standard pattern dictionary 11 that is the second dictionary.
Matching processing with the feature vector of each standard pattern in 3 is performed (step 209 in FIG. 3), and each character type category to which each standard pattern up to a predetermined rank belongs in the descending order of matching degree is a candidate for the character. The character group is output to the candidate character string buffer 108 (step 210 in FIG. 3).

More specifically, the matching unit 106
For example, a distance (Euclidean distance, Mahalanobis distance, etc.) is calculated between the character feature vector and the feature vector of each standard pattern in the standard pattern dictionary 113. Then, the matching unit 106 outputs, to the candidate character string buffer 108, the character type categories to which the standard patterns up to a predetermined rank (nth place) belong in order of increasing distance, as a candidate character group for the above-mentioned characters.

After the candidate character group buffer 108 obtains a candidate character group up to a predetermined rank in the ascending order of distance, for each of the characters constituting the character string re-cut by the character cutting unit 103, For other candidate words belonging to the candidate word group obtained in the candidate word buffer 112 for one specific character set, step 207
A series of processes 210 to 210 are repeatedly executed.

With respect to all candidate words belonging to the candidate word group obtained in the candidate word buffer 112 for one specific character set, the candidate character string buffer 108 obtains a candidate character group up to a predetermined rank for each of the characters constituting each candidate word group. When
The matching unit 106 generates a candidate character string group by combining all candidate character groups up to a predetermined rank for each character that constitutes each candidate word, and generates a candidate character string group for each candidate character string included in it. The average distance is calculated by the formula (step 212 in FIG. 3).

[0117]

(D ₁ + D ₂ + ... + D _m ) / m where m is the number of characters of the target candidate word, and D _i (1 ≦
i ≦ m) indicates the distance of the candidate character selected to form the target candidate character string at the i-th character in the target candidate word.

Then, the matching unit 106 determines that the average distance corresponding to each of the candidate character strings forming the candidate character string group from among the candidate character string groups generated corresponding to all the candidate words for one specific character set. A predetermined number (P) of candidate character strings are selected in ascending order, and they are formed into the specific character set. 2
It is output to the knowledge processing unit 114 as the recognition result of the character area sandwiched by the one specific character.

In this way, when the recognition result of the character area sandwiched by the two specific characters forming one specific character set is obtained, the process returns from step 213 to step 206 in FIG.

Then, the above-mentioned steps 201 to 2 of FIG.
From the set of candidate specific character groups obtained in the candidate character string buffer 108 by the specific character recognition process 05, another specific character set consisting of any two adjacent specific characters is extracted again, By performing the series of control processes of steps 206 to 212 of FIG. 3 again on the specific character set, the recognition result of the character region sandwiched by the two specific characters forming the specific character set is calculated. This operation is repeatedly executed for each specific character set (steps 213 → 206 to 212 → 213 in FIG. 3 are repeated).

The knowledge processing unit 114 performs knowledge processing using the entry field definition 104 and the knowledge dictionary 111 on the recognition result corresponding to each character area sandwiched between two specific characters forming each specific character set. , The final recognition result of the entire character area including the above-mentioned character areas is determined, and is output to the recognition result buffer 115 (step 21 in FIG. 4).
4).

Step 201 to FIG. 4 of FIG. 2 explained above
The final control result for each entry field is determined by repeatedly executing the series of control processing in step 214 of each step for each entry field position of the form (step 215 in FIG. 4 → repeat step 201 in FIG. 2).

In the series of recognition processes described above, reject (unrecognizable) information is added to the part of the character or character string that does not satisfy the recognition condition to the end. In this case, the recognition result obtained in the recognition result buffer 115 is displayed on the display unit 117 via the interface unit 116. When the user displays the recognition result on the display unit 117, the user uses the input unit 11 including a mouse and a keyboard.
From 8, it is possible to correct unrecognizable characters / character strings.

The user only has to specify a specific correct character in the unrecognizable character / character string from the input unit 118, and information about the correct character is output from the interface unit 116 to the correct character buffer 119 and the area coordinate buffer 120. It

In the example of FIG. 21, the image 2101 of the target character string is displayed on the display unit 117 in parallel with the recognition result 2102. When the user designates the specific area 2103 on the image 2101 with the mouse or the like which is the input unit 118, the recognition result character 2104 corresponding to it is highlighted or highlighted. In response to this display, the user inputs
When the correct character “Tou” is input from the keyboard or the like which is 18, information regarding the correct character “Tou” is output from the interface unit 116 to the correct character buffer 119 and the area coordinate buffer 120. Naturally, the user can
For example, when an area corresponding to “Tokyo” on 101 is designated and the corresponding recognition result “bunches” is corrected to “Tokyo”, information about the correct character “Tokyo” is transmitted from the interface unit 116 to the correct character buffer 119 and It is output to the area coordinate buffer 120.

The candidate word searching unit 109 uses the information about the correct character obtained in the correct character buffer 119 as the information of the specific character, and the specific character dictionary 110 and the knowledge dictionary 1 described above.
By executing the candidate word search process using 11, the unrecognizable character can be correctly recognized again.
Further, the character cutout unit 103 determines the cutout position of the correct character specified by the user as the area coordinate buffer 120.
The correct character segmentation can be performed by acquiring from.

Further, in the example of FIG. 22, the display unit 117 displays
An image of the target character string is displayed in parallel with the recognition result 2202. The user selects a specific area 220 on the image.
When 1 is designated with the mouse or the like as the input unit 118, the recognition result character 2203 corresponding thereto is highlighted or highlighted, and the recognition result candidate 2204 is displayed in the designated portion. In response to this display, when the user selects the correct answer character "Miya" from the keyboard or the like which is the input unit 118, information regarding the correct answer character "Miya" is displayed on the interface unit 11.
6 to correct character buffer 119 and area coordinate buffer 1
It is output to 20. In this case, the recognition result candidates 2204 displayed in the designated portion are the order of appearance frequency of the displayed characters, or the order of determination according to the hierarchical structure when having a hierarchical structure such as an address character string, or simply the characters. It can be configured to be displayed in code order.

As shown in FIG. 23 following the example of FIG. 22, the same correction processing is performed on the designated position 2301 and the recognition result position 2302 corresponding thereto, so that the character string 2303 is correctly recognized again. It becomes possible.

Here, regarding the re-recognition processing for each character area sandwiched between two specific characters forming each specific character set, in steps 207 to 212 in FIG. 3 described above, the characters forming one candidate word are recognized. The re-recognition process is individually executed for each of them, and the recognition result for the candidate word is finally output.

In this case, the character type searched by the matching unit 106 from the standard pattern dictionary 113 is limited to the character type of the category to which the candidate word belongs, whereby an efficient re-recognition process is realized.

On the other hand, for the entire character area sandwiched between two specific characters, the feature vector extraction and matching unit 10
The matching process according to 6 may be configured to be executed. In this case, the standard pattern dictionary 113 contains
A feature vector of a standard pattern in which each word of “Kawasaki”, “Yokohama”, “Yokosuka” ... Is set as one pattern is held, and the matching unit 106 sets a feature vector in which one candidate word is entirely set as one pattern. And the feature vector of each standard pattern in the standard pattern dictionary 113 are matched.

In this case, the word group searched by the matching unit 106 from the standard pattern dictionary 113 is limited to the word group of the category to which the candidate word belongs, whereby an efficient re-recognition process is realized.

More specifically, in recognizing an address character string, for example, the matching unit 106 sets the standard pattern dictionary 11
An efficient re-recognition process is realized by limiting the word groups searched from above 3 to the word groups constituting the hierarchical level to which the candidate word belongs.

For example, as shown in FIG. 20, in the re-recognition process of the area sandwiched between two specific characters “prefecture” and “city”, the standard pattern dictionary 113 is changed to “Kawasaki” and “Yokohama”.
It can be limited to only a word group representing a city such as “Yokosuka”.

Further, for example, in recognizing an address character string,
When the higher-level recognition result is obtained, the word group searched by the matching unit 106 from the standard pattern dictionary 113 is limited to the lower-level word group that belongs to the higher-level recognition result and to which the candidate word belongs. By doing so, a more efficient re-recognition process is realized.

For example, when the recognition result of the level 1 of the address character string is "Aomori", the standard pattern of the level 2 is
It is possible to limit not to all the word groups that can appear between the two specific characters “prefecture” and “city”, but to the word groups that represent cities that belong to “Aomori prefecture”.

Contrary to the above, when the recognition result of the lower level is obtained in the recognition of the address character string, the word group searched by the matching unit 106 from the standard pattern dictionary 113 is recognized in the lower level. By limiting the word group to which the result belongs and which constitutes the upper level to which the candidate word belongs, the unrecognizable state of the upper level can be remedied. Recording medium on which a program realizing the present embodiment is recorded
Supplement to Body The present invention is configured as a computer-readable recording medium for causing a computer to perform the same function as the function realized by each configuration of the above-described embodiments of the present invention when used by the computer. You can also do it.

In this case, as shown in FIG. 24, for example, a portable recording medium 2402 such as a floppy disk, a CD-ROM disk, an optical disk, a removable hard disk or the like, and a network line 2403 are used to realize the embodiment of the present invention. Programs for realizing various functions are stored in a memory (RAM) in the main body 2404 of the computer 2401.
(Or hard disk, etc.) 2405 and executed.

[0139]

According to the character recognition technique of the present invention, a specific character or a specific character string in an input character string is first recognized preferentially, and candidate words before and after that character string are assumed based on the recognition result. By re-recognizing the characters that make up the input character string by using the information of the candidate words, irregular intervals that are filled in various forms (slips) that are usually seen,
It is possible to recognize with high accuracy the characters that make up the input character string entered by the entry method.

According to the character correction technique of the present invention, it is possible to automatically correct other unrecognizable parts by correcting only a specific character or character string. According to the writing fluctuation control technique of the present invention, it is possible to flexibly deal with various writing methods.

[Brief description of drawings]

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

FIG. 2 is an overall control operation flowchart (No. 1) of the embodiment of the present invention.

FIG. 3 is an overall control operation flowchart (No. 2) of the embodiment of the present invention.

FIG. 4 is an overall control operation flowchart (Part 3) of the embodiment of the present invention.

FIG. 5 is a diagram showing an example of a data format of an entry field definition.

FIG. 6 is a control operation flowchart (No. 1) of the character cutout unit.

FIG. 7 is a control operation flowchart (No. 2) of the character cutout unit.

FIG. 8 is an explanatory diagram of a control operation of a character cutout unit.

FIG. 9 is a control operation flowchart for fluctuation in notation.

FIG. 10 is a structural diagram of a knowledge dictionary (address).

FIG. 11 is a structural diagram of a specific character dictionary.

FIG. 12 is a diagram showing a structural example (part 1) of a specific character dictionary 110.

FIG. 13 is a diagram showing a structural example (part 2) of a specific character dictionary 110.

FIG. 14 is a diagram showing a structural example (part 3) of a specific character dictionary 110.

FIG. 15 is a diagram showing a structural example (part 4) of a specific character dictionary 110.

FIG. 16 is an explanatory diagram of a fluctuation control operation of the notation.

FIG. 17 is an operation explanatory diagram (1) of the candidate word search unit.

FIG. 18 is an explanatory diagram (No. 2) of operation of the candidate word search unit.

FIG. 19 is a diagram for explaining the operation of the candidate word search unit (No. 3).

FIG. 20 is an explanatory diagram of a character string detection / recognition operation using a standard pattern dictionary.

FIG. 21 is an operation explanatory diagram (1) of the input unit and the display unit.

FIG. 22 is an operation explanatory diagram (2) of the input unit and the display unit.

FIG. 23 is an operation explanatory diagram (3) of the input unit and the display unit.

FIG. 24 is an explanatory diagram of a recording medium in which a program that realizes this embodiment is recorded.

[Explanation of symbols]

101 Input string 102 image memory 103 character cutout section 104 Entry field definition 105 Feature Extraction Unit 106 Matching unit 107 Specific Character Standard Pattern Dictionary 108 Candidate character string buffer 109 Candidate word search unit 110 Specific character dictionary 111 Knowledge Dictionary 112 candidate word buffer 113 Standard Pattern Dictionary 114 Knowledge Processing Unit 115 Recognition result buffer 116 Interface part 117 Display 118 Input section 119 Correct character buffer 120 area coordinate buffer

Continuation of front page (56) Reference JP-A-7-262320 (JP, A) JP-A-3-257693 (JP, A) JP-A-5-6464 (JP, A) JP-A-6-4717 (JP , A) JP-A-2-101596 (JP, A) JP-A-8-171614 (JP, A) JP-A-5-89291 (JP, A) PRU93-49 Study on post-recognition processing method of address character string , IEICE Technical Report, Japan, September 16, 1993, Vol. 93, No. 228, pp. 49-56 Handwriting based on key character driven place name inference, address recognition, IEICE Transactions, Japan, May 1997, Vol. J80-D-II No. 5, pp. 1077-1085 (58) Fields investigated (Int.Cl. ⁷ , DB name) G06K 9/00-9/82

Claims (4)

(57) [Claims] 1. A character recognition method for recognizing a character constituting an input character string entered in an entry field having a predetermined category, which corresponds to the input character string and a specific character or a specific character string.
By executing a first matching process with a first recognition dictionary that stores a standard pattern, a specific character or a specific character string is extracted from the input character string, and belongs to the predetermined category, and the input Specific character or specific character string and two specific sentences extracted from the character string
A candidate word group that may be located in a region in the input character string between characters or a specific character string is extracted from the category word dictionary, and for each candidate word belonging to the extracted candidate word group, respective candidate words are included based on information about each candidate word
That for each region in the input character string, the candidate word group
Corresponding to a character or character string related to a candidate word belonging to
A character recognition method, comprising: recognizing a character forming the input character string by executing a second matching process using a second recognition dictionary storing a standard pattern . 2. A character correction method using the character recognition method according to claim 1 , wherein a recognition result of characters forming the input character string is displayed in parallel with the input character string, and the displayed result is displayed. A desired area on the input character string is designated to correct the character or character string corresponding to the desired area, and the candidate word group is extracted based on the correct character or the information on the correct character string given by the correction. A character correction method comprising: recognizing a character forming the input character string again by re-executing the processing and the second matching processing. 3. The method according to claim 2 , further comprising: displaying a plurality of candidate recognition results in the desired area in response to designation of the desired area on the displayed input character string. Character correction method characterized by the following. 4. A recording medium for recording a program read by the computer when it is used by the computer, which corresponds to an input character string entered in an entry field having a predetermined category and a specific character or a specific character string. Standard power
A function of extracting a specific character or a specific character string from the input character string by executing a first matching process with a first recognition dictionary storing a turn , and belonging to the predetermined category, and Specific character or specific character string and two specific sentences extracted from the input character string
A function of extracting a candidate word group that may be located in a region in the input character string between letters or a specific character string from a category word dictionary, and for each candidate word belonging to the extracted candidate word group , The candidate word group for each area in the input character string that includes each candidate word based on information about each candidate word
Corresponding to a character or character string related to a candidate word belonging to
By executing a second matching process using a second recognition dictionary that stores a standard pattern, a function for recognizing the characters that form the input character string and a program for causing the computer to perform are recorded. Computer readable recording medium.