JPH02268389A - Post processor for character reader - Google Patents

Abstract

PURPOSE:To improve the correction efficiency of an output candidate character string by replacing the output candidate character with an erroneously recognized character or inserting a dummy character as the lack part of a character number and executing knowledge processing. CONSTITUTION:Recognition processing is executed by a character read part 11 concerning the character written to a slip as the object of character read processing. The obtained output candidate character is transferred to a knowledge processing part 12 and held in a candidate character storage part 13. A comparison part 15 collates each output candidate character string at every held word with a character string group, which is stored in a knowledge dictionary storage part 14, and selects the character string of highest similarity. Then, the data of the character string converted by a character conversion part 16 are displayed in a display part 18. When the displayed character string is not proper as a final character read result, the displayed character string is replaced with the erroneously recognized character or the specified character is inserted as the lack part of the character number from an input part 19 and the knowledge processing of the output candidate character string is executed. Thus, the correction efficiency of the output candidate character string can be improved.

Description

[Detailed Description of the Invention] [Object of the Invention] (Industrial Application Field) The present invention compares an output candidate character string obtained as a result of character recognition processing with a group of character strings stored in a dictionary. Accordingly, the present invention relates to a post-processing device for a character reading device that outputs a final character reading result.

(Prior Art) Generally, recognition results (output candidate characters) obtained by character recognition processing in an optical character reading device are subjected to post-processing called knowledge processing to determine the final character reading result. In knowledge processing, for example, when the contents of character strings to be entered in a form are limited in advance, such as place names, personal names, corporate names, etc., a knowledge dictionary containing these character string groups is prepared. Then, the final character reading result is determined by comparing the character string group stored in this Chichu dictionary with the output candidate character string.

Normally, in knowledge processing, the number of characters in an output candidate character string is made to correspond to the number of characters in a character string to be compared that is stored in a dictionary. For this reason, if the number of characters in the output candidate character string is insufficient due to an error in filling out a form, etc., matching with the character string that should originally be used for knowledge processing is not performed.
Unable to obtain correct character reading results. In such a case, the output candidate character string is corrected by inputting correct character data from a keyboard or the like, and knowledge processing is performed again to obtain a correct character reading result.

(Problem to be Solved by the Invention) By the way, if the number of characters in the output candidate character string is insufficient, or if the output candidate character string is not matched with the corresponding character string group, erroneously recognized characters may be included in the output candidate character string. Corrections that were made when the words were included in the text were performed by inputting the phonetic phrase for each character. For this reason, the efficiency of correcting output candidate character strings was low.

The present invention has been made in view of the above points, and an object of the present invention is to provide a post-processing device for a character reading device that can improve the efficiency of modifying output candidate character strings.

[Structure of the Invention] (Means for Solving the Problems) The present invention provides for, when the output candidate character string obtained through recognition processing is not valid as the final character reading result, the output candidate character string is insufficient. an input means for manually inputting a specific character indicating that the character is not a valid character corresponding to a character that has been misrecognized or a character that has been misrecognized; and a setting means for setting the output candidate character string by inserting it into the character position of the character that has been misrecognized or replacing it with the erroneously recognized character, and the output candidate character string set by the setting means. Characters other than the specific characters in
The apparatus is configured to compare characters in a character string stored in the dictionary that correspond to the same character, and determine a final character reading result based on the result of the comparison.

(Operation) In the knowledge processing device of the character reading device configured in this way, a specific character indicating that the output candidate character string is not a valid character is added to the output candidate character string. A comparison is made with a group of character strings stored in a dictionary corresponding to .

Further, by adding the specific character in place of the incorrectly recognized character in the output candidate character string, the final character reading result is determined based on the correctly recognized characters other than the specific character. I can do it.

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing the configuration of a knowledge processing device of a character reading device according to the same embodiment. In the figure, a character reading section 11 performs recognition processing on characters written on a form, and is connected to a knowledge processing section 12 that performs knowledge processing on output candidate characters obtained through the recognition processing. The knowledge processing unit 12 includes a candidate character storage unit 13 and a knowledge dictionary storage unit 14.
, comparison section 15, character conversion section 18. and a display control section 17. The candidate character storage section 13 is connected to the character reading section 11 and holds output candidate character strings obtained as a result of recognition processing in the character reading section 11. The knowledge dictionary storage unit 15 stores a group of character strings to be compared when determining the validity of output candidate character strings. Also,
The candidate character storage section 13 and the knowledge dictionary storage section 15 are connected to the comparison section 14. The comparison unit 14 compares the output candidate character string stored in the candidate character storage unit 13 with the knowledge dictionary storage unit 1.
4 is compared with the string group stored in 4. Comparison section 15
A character conversion unit 1B is connected to the character conversion unit 1B, which replaces characters in the output candidate character string with other characters or dummy characters (described later). A display control section 17 that controls display of final character reading results and the like is connected to the character conversion section 1B. A display section 18 is connected to the display control section 17 . Further, an input unit 19 configured by, for example, a keyboard is connected to the character conversion unit 1G, for inputting corrected characters or dummy characters for the output candidate character string.

Next, the operation of this embodiment will be explained.

First, the character reading section 11 performs recognition processing for each character written on a form to be subjected to character reading processing. In the form to be processed, a character entry area is set in advance for each word (character string) to be entered. The character reading unit 11 stores, for example, the top 10 recognition results (output candidate characters) in the candidate character storage unit 13 for each character written in the character entry area obtained by the character recognition process.
Transfer to.

When the knowledge processing section 12 receives the recognition results (output candidate characters) transferred from the character reading section 1, it performs knowledge processing according to the flowchart shown in FIG.

Here, it is assumed that the character string entered in the form is the address name ``Kanagawa Prefecture, Yokohama, Line District, Ukikaoka'' as shown in Figure 3. The character string shown in FIG. 3 shows an example in which the character "city" was not entered by mistake when "Yokohama type" should have been entered. Also, the character strings entered in the form may be changed to related words such as address names (
"God Festival) 1 Prefecture""YokohamaType""LineSection""BeautifulHill"
), it is assumed that the knowledge dictionary set in the knowledge dictionary storage unit 14 has a hierarchical structure in order to obtain reliable recognition results. In other words, if the reading result "Kanagawa Prefecture" is obtained through knowledge processing, the character string group (knowledge dictionary) corresponding to "Kanagawa Prefecture" is identified, and the knowledge processing is performed for the second word ("Yokohama type"). processing is performed. Similarly, when the reading result of "Yokohama type" is obtained through the knowledge processing, the knowledge dictionary to be used when performing the knowledge processing on the third word ("line section") is specified. The same goes for the fourth word (``Utsukushigaoka''). Furthermore, the knowledge dictionary used in the knowledge processing is the same as the number of characters in the output candidate character string to be processed.

First, when the candidate character storage section 13 receives the recognition results (output candidate characters) of the characters written on the form, it holds the output candidate characters for each word (step Sl). The first output candidate character string stored in the candidate character storage unit 13 is
Shown in the figure. As shown in FIG. 4, in the output candidate character string for which knowledge processing has not been performed, five characters indicated by underlines in the figure are erroneously recognized. When the output candidate character strings are stored in the candidate character storage unit ta, the comparison unit 15 compares each output candidate character string for each word stored in the candidate character storage unit 13 with each output candidate character string stored in the knowledge dictionary storage unit 15. Check the knowledge dictionary (character string group) corresponding to the output candidate character string (word) (
Step 32). First, the comparison unit 14 compares the 10th character string “Kanagawa Prefecture” stored in the candidate character storage unit 13.
The character strings with the highest similarity are compared with the knowledge dictionary (corresponding to 4 characters) for the first word stored in the knowledge dictionary storage unit 15, and the character strings with the highest similarity are selected. Select from knowledge dictionary. The character conversion unit 16 is
If there is a character in the first output candidate character string that is different from a character in the character string selected from the knowledge dictionary, this different character is converted into a corresponding character in the character string selected from the knowledge dictionary. For the first word, the second character "ho" that was incorrectly recognized is converted to the correct character "na". next,
Knowledge processing regarding the second word "Yokohama" is performed. 2
The second word "Yokohama" is erroneously recognized as "Bakuto" as shown in FIG. Normally, the second word should have been written as ``Yokohama type'' with 3 characters, but only 2 characters were written, so knowledge processing cannot restore the correct character string. That is, the knowledge dictionary storage unit 1
The lower level knowledge dictionary (knowledge dictionary for the second word) corresponding to the character string "Kanagawa Prefecture" set in Here, it is assumed that the first output candidate character ``Bukuro'' for the second word remains as a result of knowledge processing.In this way, knowledge processing is performed correctly at the upper level. If this is not the case, it will not be possible to correctly perform knowledge processing on the third and fourth words because the knowledge dictionary has a hierarchical structure.

As a result, it is assumed that the first knowledge processing yields a character string as shown in FIG.

The character string data obtained by conversion by the character conversion section 16 is output to the display control section 17. Display control section 17
displays a character string as shown in FIG. 5 on the display section 18 based on the character string data from the character conversion section te (step S3).

When the character reading result is displayed on the display unit 18, a character string correction mode is entered for correcting the reading result. here,
When the operator determines that there are no misrecognized characters in the character string and is notified that the character string does not need to be corrected, the displayed character string is determined as the final character reading result. (Step S4). On the other hand, if there are erroneously recognized characters in the displayed character string, character data including dummy characters for correcting the character string is input by the operator from the input unit 19 (step S5).
. As the dummy character, for example, a specific character such as the symbol "x" is set in advance. These dummy characters can be inserted into the character string displayed on the display unit 18, for example, in the character position corresponding to the missing character when the number of characters is insufficient, or can be inserted in place of a character that has been misrecognized. can be entered.

In addition to inputting dummy characters from the input unit 19,
It is possible to input the phonetic lessons of normal character data. The character conversion unit 1G sets output candidate characters by inserting them into the output candidate character string at positions corresponding to missing characters or by replacing them with erroneously recognized characters according to the input character data (step S6).

Here, as shown in Figure 6, for the reading result of the second word "Bakufu", the first character is converted to the correct character "Yoko" by phonetic input, and the second character is changed to a dummy character. By converting and adding (retrofitting) one dummy character, it is assumed that the character string "Horizontal XXJ" is set.

When knowledge processing is activated after the character string is converted into a character string with dummy characters added in this way, the character conversion unit 1B transfers the character string data to the comparison unit 14. When the comparison unit I4 receives the character string data, it performs knowledge processing on the character string to which the dummy characters are added (step S2). Comparison section 14
performs knowledge processing by treating the dummy character as one character. That is, the comparison unit 14 performs a comparison with the knowledge dictionary corresponding to the three characters stored in the knowledge dictionary storage unit 15 for the second word “Horizontal xxJ”.
It is not a target when calculating the degree of similarity with the knowledge dictionary in comparison fi15. Therefore, knowledge processing regarding the character string "horizontal XX" is performed based on the degree of similarity regarding the first character "horizontal". Knowledge dictionary storage! 5 stored in “
Lower level knowledge words corresponding to “Kanagawa-kei”! F (3-letter knowledge dictionary for the second word) contains "Yokohama type"
It includes a dictionary. Therefore, the character string "Horizontal XXJ" is
Since this knowledge dictionary has the highest degree of similarity with the knowledge dictionary "Yokohama type", this knowledge dictionary is selected by the comparison unit 15. However, if the knowledge dictionary contains multiple 3-character string dictionaries whose first character is ``yoko'' other than ``Yokohama type'', ``Dictionaries other than Yokohama city j'' may be selected. In such a case, candidate character strings are sequentially displayed in response to pressing of the next candidate key, for example.

When the correct character string is selected by the operator from the character strings sequentially displayed, this can be determined as the character reading result. In this way, by converting the second word into a correct character string through knowledge processing, the knowledge dictionaries corresponding to the words in the upper level hierarchy are also specified for the third and fourth words. Therefore, the third and fourth words are also converted into correct character strings through knowledge processing. The character string thus obtained is shown in FIG.

In this way, there is no need to correct all the misidentified characters by inputting the pronunciation for each character, and knowledge processing can be performed by inputting specific characters designated as dummy characters, making the operation easy. This can improve work efficiency. In addition, even if the number of characters entered in a form differs from the original number due to an entry error, by inserting dummy characters, it is possible to compare it with a predetermined group of character strings that should be used for the original knowledge processing. It is done.

[Effects of the Invention] As described above, according to the present invention, knowledge processing can be performed by replacing erroneously recognized characters with respect to output candidate characters or by inserting dummy characters to fill in the insufficient number of characters. Even when the number of characters in a written character string is insufficient, knowledge processing can be performed reliably and output candidate characters can be corrected efficiently.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a knowledge processing device of a character reading device according to an embodiment of the present invention, FIG. 2 is a flowchart showing a processing procedure when performing knowledge processing in the same embodiment, and FIGS. FIG. 7 is a diagram for explaining the operation of the same embodiment. 11...Character reading section, 12...Knowledge processing section, 13.
... Candidate character storage section, 14... Knowledge dictionary storage section, 15
... Comparison section, 16... Character conversion section, 17... Display control section, 18... Display section, 19... Human power section. Diagram

Claims (1)

[Claims] An output candidate character string obtained by performing recognition processing on characters written in a form that is subject to character reading processing, and a comparison target when determining the validity of the output candidate character string. Based on the dictionary that stores the strings,
In the post-processing device of the character reading device that determines the final character reading result, if the output candidate character string is not valid as the final character reading result, the missing characters in the output candidate character string, or an input means for inputting a specific character indicating that the incorrectly recognized character is not a valid character; a setting means for setting the output candidate character string by inserting it at a character position or replacing the erroneously recognized character; and a character other than the specific character in the output candidate character string set by the setting means. and post-processing means for collating the characters in the character string stored in the dictionary that correspond to the same character, and determining a final character reading result based on the collation result. A post-processing device for character reading devices.