JP2968354B2 - Post-processing method of character recognition result - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a post-processing method for correcting an error in a character recognition system by word collation or morphological analysis in a character recognition system.

[0002]

2. Description of the Related Art In a character recognition system, it is extremely difficult to completely eliminate erroneous recognition by character recognition on a character-by-character basis using a characteristic amount of a character image. For this reason, in a Japanese OCR system that recognizes printed characters or handwritten characters, a character string obtained as a result of character recognition often automatically corrects an erroneously recognized character by word matching or morphological analysis in post-processing. .

Many post-processing methods for such a Japanese OCR system have been proposed, but the following methods are known for general sentences. Note that c and d
Is a method in which a patent application is filed by the same application. a) After the recognition of the area to be recognized (for example, a paragraph) is completed, post-processing is performed on all character strings in the area in order from the beginning, assuming that there is no line break. b) A sentence is used as a unit of processing, and a character string from the beginning of the sentence to a period or a punctuation mark is processed ignoring a line feed. c) The unit of processing is a line, the character string of the line is separated by the change position of the character type, the character of the last delimiter is moved to the beginning of the next line, and the next line is processed. d) The unit of processing is a line, word processing is performed from the beginning of the line, characters at the end of the line remaining unanalyzable are moved to the beginning of the next line, and processed in the next line.

Examples of known materials relating to post-processing of character recognition results of Japanese documents by word collation and morphological analysis include, for example, “Nishino et al .:“ Realization of Post-processing of Japanese Reader ””, Natural Language Processing 64-6 (1987). .11.2.
0), pp. 45-52 ".

[0005]

According to the above method a or b, when a paragraph or a sentence is long, the start of post-processing is waited until the recognition is completed, so that the processing efficiency is low. Therefore, there is a problem that a large memory capacity is required.

A character string is analyzed by being delimited by words or phrases, but one word may be split at the end of a line and at the beginning of the next line. When performing post-processing on a line-by-line basis, unless such word splitting is taken into account, the split word cannot be correctly analyzed, and correction of misrecognized characters in the split word fails.

[0007] In the above-mentioned method c or d using a line as a processing unit, the word split from the end of the line to the head of the next line is processed in the next line, so that the error of the split word is corrected. It is possible. In addition, since the unit is a line, at least a memory for storing the recognition result until the post-processing is completed,
Recognition processing of the next row can be executed in parallel with post-processing of a certain row. However, there are still issues to be solved.

That is, in the above method c, since the end of the line is determined by the delimiter based on the change of the character type, if all the characters in the line are hiragana or alphanumeric, the character type information alone is used. A column break cannot be specified, and all characters in a line are moved to the next line. That is, there is no guarantee that the number of characters transferred to the next line is not necessarily small. In order to cope with such a situation, it is necessary to provide a margin in a memory for storing the recognition result waiting for post-processing, and there has been a problem that the purpose of memory saving cannot be sufficiently achieved.

In the method d, the word analysis becomes impossible to determine that the character string at the end of the line is to be transferred to the next line. However, when word matching is performed, the end of the line may be split. Words are not always unparsable. Some broken words at the end of a line may be successfully matched with another word that is not the correct answer. In this case, since the analysis is not disabled, the transfer to the next line is not performed, and there is a problem that an erroneous correction is determined.

[0010] An object of the present invention is to solve the problems of the conventional method, and it is possible to reliably correct a word broken by a line feed.
Another object of the present invention is to provide a post-processing method capable of sufficiently reducing the capacity of a memory for storing necessary information until the end of the post-processing even when the character recognition processing and the post-processing are performed in parallel.

[0011]

According to the post-processing method of the present invention, the processing is executed for each line in units of a word from the head of the line, and each time the processing of one word is completed, the unprocessed word following the word is processed. The number of characters to be processed is compared with a certain value. If the number of characters of the unprocessed character is equal to or less than the certain value, the processing of the line being processed is terminated, and the processing is performed in addition to the unprocessed character or the unprocessed character. The character of the last word part that has already been completed is also moved to the beginning of the next line.

In each line, when the number of characters excluding characters transferred from the previous line is less than a certain value, or
In each line in the processing target area, when the number of characters excluding the characters transferred from the previous line is smaller than the maximum number of characters of the processed line by a certain value or more, or when the number of characters is equal to or less than a certain ratio of the maximum number of characters, and If the last character is a punctuation mark, process the line up to the last character.

[0013]

Since the post-processing is performed for each line, post-processing is executed every time the character recognition processing for one line is completed, thereby improving the processing efficiency. Also, word analysis is performed as a continuous character string in the next line, so that it can be appropriately corrected.

By moving the character string of the last processed word in addition to the unprocessed character string to the next line, even in the case where the inter-word connection is verified by the word analysis processing, the word of the word split by the line feed can be used. It is possible to verify the inter-connection, and a reliable analysis is guaranteed.

Since the maximum number of characters of the character string transferred to the next line does not exceed a certain value, the amount of memory required for storing the character string transferred to the next line can be extremely reduced.

In addition, the condition determination prevents a character string from being carried over to the next meaningless line, so that wasteful processing can be reduced and efficient post-processing can be performed.

[0017]

FIG. 1 is a schematic block diagram of a Japanese character recognition system according to the present invention. In this Japanese character recognition system, an image input unit 10 reads an image of a document original by a scanner or the like, and stores the binary image data in a document image memory 11. The line / character extracting unit 12 extracts a character line and a character image from the image in the document image memory 11, stores the character image data in the character image memory 13, and recognizes a recognition target area (for example, a paragraph. , Specified by the operator), and the cutout information such as the position of the cutout line, the character position, and the character size are stored in the cutout information memory 14.

The character recognizing unit 15 reads out character image data from the character image memory 13, performs normalization, extracts a characteristic amount, and performs matching between the characteristic amount and a dictionary in the character dictionary memory 16. Further, the recognition result candidate is determined by using the shape information in the cutout information, and the character code and the distance data thereof are stored in the recognition result memory 17.

The post-processing section 18 is a section for executing post-processing directly related to the present invention. That is, after reading the recognition result line by line from the recognition result memory 17 and adding a character string carried over to the previous line from the line end storage memory 19 at the beginning,
From the beginning of the line, language analysis / correction processing is performed using the contents of the word dictionary memory 20, and the recognition result memory 1
7 is rewritten. When it is necessary to move the character string at the end of the processing line to the next line, the character string is written to the end-of-line storage memory 19. Details of the contents of this post-processing will be described later for each embodiment.

The result output unit 21 outputs the contents of the recognition result memory 17 to an output device such as a display or a printer or a file device such as a magnetic disk device.

Embodiment 1 The contents of the post-processing will be described in order along the flowchart shown in FIG. First, the post-processing unit 18 reads, from the recognition result memory 17, recognition result data of one line (to be referred to as a current line) currently being processed (step 100).

Next, it is checked whether there is a character string carried over from the previous line in the line end storage memory 19 (step 102). If there is, the character string is read from the line end storage memory 19 and the recognition result data of the current line is read. (Step 104).

Next, the recognition result data of the current line (including the character string carried over from the previous line) is analyzed by word matching one word at a time from the beginning of the line (step 106).

Each time one word is processed, it is checked whether or not the number of characters of the unprocessed character string following the processed word is equal to or less than a certain value (threshold for the number of characters at the end of a line) (step 10).
8). This is one step of judging whether the unprocessed character string is carried over to the next line. If the number of unprocessed characters is larger than the threshold value of the number of characters at the end of the line, it means that the process has not progressed to the stage of carrying over to the next line, and the process proceeds to the next word.
The threshold value for the number of characters at the end of the line may differ depending on the system, but a value smaller by one than the number of characters having the maximum word length in the word dictionary used in the post-processing unit 18 is used. Here, the description will proceed assuming that the maximum word length is 10 characters and the line end character number threshold is 9.

If the number of unprocessed characters is equal to or smaller than the line end character number threshold, it is checked whether the number of unprocessed characters is 0 (step 1).
10). If the number of unprocessed characters is 0, the processing of the current line has been completed up to the last character, so it is checked whether or not the next line remains (step 112). If not, that is, the current line is the last line of the recognition target area. If so, the process is completed. If there is a next line, the process returns to step 100 to start the processing of the next line.

If it is determined in step 110 that unprocessed characters remain, it is checked whether the last character is a punctuation mark or a punctuation mark (step 114). If it is a punctuation mark, it is determined that the current line should be processed to the end. Therefore, the process returns to step 106 to execute the processing of the next word.

When the last character is not a punctuation mark, the number of characters excluding the carryover character from the previous line of the current line is smaller than the maximum number of characters in the processed line in the recognition target area by a certain threshold value or less. It is checked whether the number is equal to or less than a predetermined ratio of the maximum number of characters (step 116). If such a condition is not satisfied, it is determined that the carry-over to the next line is not performed at this time, and the process returns to step 106. On the other hand, if the condition is satisfied, it is checked whether or not the number of characters in the current line excluding the number of characters carried over from the previous line is equal to or smaller than the line end character number threshold (step 118). If the condition is not satisfied, step 106
Return to

If the condition of step 118 is not satisfied, it is checked whether the current line is the last line of the recognition target area (step 120). If it is the last line, the current line is not carried over to the last character of the current line. Step 1 for processing on the current line
Returning to step 06, if it is not the last line, it is finally determined that the unprocessed character string of the current line is carried over to the next line, and this character string is stored in the end-of-line storage memory 19 (step 122). The process ends, and the process returns to step 100 to start the process of the next line.

Here, the input and display of the next two lines of character string images will be demonstrated. The post-processing will be described using a recognition target area composed of.

When the analysis of the character string on the first line is performed by word collation from the beginning, the processing proceeds as follows up to the fourth word. / Input / and / of image / (however, / means word boundary)

At this point, since the number of remaining characters on the first line is 10, the process proceeds to the next word. At the stage where the next word “display” is processed, the number of unprocessed characters becomes 8, which is 9 or less, which is the line end character number threshold value. Demonstration of image /// input / and / display / 8 7 6 5 4 3 2 1 (Unprocessed character strings and number of characters are underlined)

The last character is not a punctuation mark and the current line is the first line, so that the condition of step 116 is not naturally satisfied. Further, the number of characters of the current line is 16, which is more than 9 which is the threshold of the number of characters at the end of the line. Line is not the last line. Therefore, at this stage, the processing of the current line, that is, the first line, is completed.
Unprocessed character string "Demonstration" is stored at the end of line memory 1.
9 and proceed to the processing of the second line. Here, since the line end character number threshold value is 9, it is sufficient that the line end storage memory 19 has a capacity of 9 characters.

When the character string of the recognition result on the second line is read and the content of the line end storage memory 19 is added to the beginning of the character string, the next character string is obtained. That is, the word “demonstration” split at the first line feed has a continuous form. Conduct a demonstration. (The underlined part is the previous line carryover character string)

As the analysis of this character string proceeds, the following is demonstrated: / demonstration / perform / perform. become that way. However, a) The end-of-line character is a period (step 114). b) The number of characters excluding the carry-over character is included, including punctuation.
8, the number of characters at the end of the line is smaller by 9 or more than the maximum number of characters 16 of the processed line, and the ratio to the maximum number of characters is 8/16 =
0.5, which is sufficiently small (step 116). c) The number of characters excluding the carry-over character of the preceding line is equal to or smaller than the line end character number threshold (step 118). d) The last line (step 120). Therefore, the last character is processed and is not carried over to the next line.

Embodiment 2 In step 122 of FIG. 2, the character string of the last processed word is stored in the end-of-line storage memory 19 in addition to the unprocessed character string, and is carried over to the next line. The content is different.

The two-line character string used in the first embodiment is used again as an example. The processing proceeds as follows up to the word “display” on the first line. Demonstration of images / input / and / display /

At this stage, the processing on the first line is completed, and the character string after the word "display" processed last is carried over to the next line.
Therefore, the second line performs a demonstration of the display. If word analysis processing is performed, display / no / demonstration / do / do /. Becomes

Here, the reason why not only the unprocessed word but also the character string of the last processed word is carried over to the next line is as follows. If the word analysis process in post-processing is to verify not only word matching but also connections between consecutive words, if only unprocessed character strings are carried over to the next line, The inability to verify the connection between the word and the word immediately before it (the word processed last in the previous line) leaves insufficient analysis. Such inconvenience can be avoided by carrying over the last processed word to the next line as in the present embodiment.

A description will be given again using the two lines of character strings given in the example. According to the first embodiment, the “demonstration” is carried over to the beginning of the second line, and the word “no” is processed first, but the information of “display” preceding it is not carried over. Cannot verify the connection between "Display" and "No". On the other hand, according to the present embodiment, “display” is also 2
Since the data is carried over to the second line, there is no discontinuous point in connection verification between “display” and “no” in the processing of the second line.

In other words, the end-of-line processing may be selected according to the level of the word analysis processing in the post-processing. If the analysis is only for word matching, there is no problem in the end-of-line processing of the first embodiment. If the analysis also performs inter-connection verification, the line ending processing of the second embodiment is appropriate.

The post-processing section 19 for executing the above-described processing is realized by software using a general-purpose processor, or is realized by using dedicated hardware. In any case, a person skilled in the art can easily realize the post-processing unit 19 based on the above description.
The illustration of specific examples of software or hardware for implementing the above is omitted.

[0042]

As described in detail above, according to the present invention, the following effects can be obtained. 1) Since the post-processing is performed for each line, the post-processing is executed every time the character recognition processing for one line is completed, thereby increasing the processing efficiency. 2) End-of-line processing, that is, carry-over processing of the unprocessed character string or the last processed word and the unprocessed character string to the next line. The word can be analyzed and corrected appropriately. 3) By carrying over not only the unprocessed character string but also the final processed word to the next line in the end-of-line processing, even when verifying the inter-word connection by the word analysis processing, the verification of the inter-word connection of a word split by a line feed can be performed. It is possible and a reliable analysis is guaranteed. 4) Comparison of the number of unprocessed characters, the number of characters excluding carryover characters from the previous line, or the number of characters excluding carryover characters from the previous line with the maximum number of characters in the processed line, and judgment of conditions such as the type of the last character Thus, unnecessary post-processing can be prevented and unnecessary post-processing can be performed by reducing waste of processing. 5) Since the maximum number of characters of the character string carried over to the next line does not exceed a certain value, the amount of memory required for storing the character string transferred to the next line can be extremely reduced.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram of a Japanese character recognition system according to the present invention.

FIG. 2 is a flowchart of post-processing according to an embodiment of the present invention.

[Explanation of symbols]

 Reference Signs List 10 Image input unit 11 Document image memory 12 Line / character extraction unit 13 Character image memory 14 Extraction information memory 15 Character recognition unit 16 Character dictionary memory 17 Recognition result memory 18 Post-processing unit 19 End-of-line storage memory 20 Word dictionary memory 21 Result output unit

Claims (5)

(57) [Claims] 1. In post-processing for correcting an error in a character recognition result by a word analysis process or the like, a process is executed for each line in a word unit from a line head, and each time one word process is completed, the process is performed. The number of characters of the unprocessed characters following the word is compared with a certain value. If the number of characters of the unprocessed characters is equal to or less than the certain value, the processing of the line being processed is terminated, and the unprocessed characters are replaced with the beginning of the next line. Post-processing method of character recognition result, characterized in that it is transferred to 2. In post-processing for correcting an error in a character recognition result by a word analysis process or the like, a process is executed for each line in units of a word in order from the head of the line, and each time the processing of one word is completed, the process is performed. The number of unprocessed characters following the word is compared with a certain value. If the number of unprocessed characters is equal to or less than the certain value, the processing of the line being processed is terminated. A character recognition result post-processing method, which moves characters up to the beginning of the next line. 3. In each line, when the number of characters excluding characters transferred from the previous line is equal to or less than a certain value, processing is performed on the line up to the last character.
Post-processing method of the described character recognition result. 4. In each line in the processing target area, when the number of characters excluding characters transferred from the previous line is smaller than a maximum number of characters of a processed line by a certain value or more, or when the number of characters is less than a certain ratio of the maximum number of characters. 3. The post-processing method according to claim 1, wherein the last character is processed in the line. 5. The post-processing method according to claim 1, wherein, when the last character in each line is a punctuation mark, processing is performed on the line up to the punctuation mark.