JPS60254387A - Character correction method in character recognition device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Technical Field of the Invention] The present invention provides a character correction method for correcting a misrecognized character when a character recognized by a character recognition device, such as a kanji character, is incorrect. Regarding.

[Technology background]

In general, character recognition devices optically observe characters written on paper, etc., extract the characteristics of the characters, and then compare the characteristics of the characters with the contents of a predetermined recognition dictionary to identify the characters. The recognition result is displayed on the display unit. However, with such a character recognition device, there is a risk that the characters may not be recognized correctly, especially when the characters written on paper or the like are disordered or have curly characters. Therefore, it is necessary to correct the displayed characters that are incorrectly recognized.

[Conventional technology and problems]

Conventionally, in this type of character correction method, for example, a plurality of candidate character groups including recognized characters are selected in advance from a recognition dictionary in order of similarity to observed characters, and the candidate character groups are displayed on a display. One that displays the characters in sequence and selects the correct one to correct the misrecognized characters.
By inputting information to specify the correct character, such as the kana reading and number of strokes of the correct character, on a keyboard device, a group of characters for correction is selected from a predetermined dictionary based on the key input. There is a system in which a group of correction characters are sequentially displayed on a display section, and a correct one is selected from among them to correct a misrecognized character.

However, in such conventional character correction methods,
In the former case, since the candidate character group has a high degree of similarity to the observed character, the correct character is not necessarily included in the candidate character group. If the correct character is not included in the candidate character group, a problem arises in that the method cannot correct the misrecognized character. In addition, in the latter case, the correct character is always included in the correct character group, so the problem like the former does not occur, but for example, when the number of characters in the correction character group is large, This poses a problem in that it takes time to select the correct character from a group of characters for correction, and the speed of character correction becomes slow accordingly.

[Purpose of the invention]

This invention was made based on the above points,
The purpose is to provide a character correction method in a character recognition device that allows incorrectly recognized characters to be corrected quickly and accurately at all times.

[Structure of the invention]

In order to achieve the above-mentioned object, the present invention selects from a recognition dictionary a plurality of candidate characters ranked in descending order of similarity to the observed character, and selects the first-ranked candidate character from the group of candidate characters. It is based on a character recognition device that displays the recognition result on the display, and focuses on the relatively high probability that the correct character will be included in the candidate character group, and uses it to correct incorrectly recognized displayed characters. , a correction character group including the correct character is selected from a predetermined dictionary based on information for specifying the correct character, and if the candidate character is included in the correction character group, the corresponding After moving the correction character that matches the candidate character to a higher position than other correction character groups according to the ranking of the candidate character group, select the correction character that matches the correct character from among the correction character group with this new arrangement order. The correction character is selected and the correction character is redisplayed on the display section in place of the erroneously recognized character.

[Embodiments of the invention]

Hereinafter, the present invention will be described in detail based on the accompanying drawings.

FIG. 1 is a block diagram showing an embodiment of a character recognition device for realizing the character correction method according to the present invention. In the figure, 1 is an observation unit that optically observes characters such as kanji written on paper, etc., 2 is a feature extraction unit that extracts features of input characters based on observation data from observation unit 1, and 3 is 4 is a recognition dictionary that stores standard feature patterns of characters to be recognized in advance, and 4 is a comparison between the feature pattern of the input character transferred from the feature extraction unit 2 and the standard feature pattern in the recognition dictionary 3. a collation identification unit that identifies characters in the recognition dictionary 3 as input characters in descending order of similarity;
5 is a candidate storage memory that stores each candidate character (for example, 5 characters) identified by this collation identification unit 4 in descending order of similarity;
6 is a kana-kanji conversion dictionary that stores in advance correction character groups corresponding to each kana input and the number of the correction character groups; 7
8 is a display unit; 9 is a candidate storage memory 5;
a display output section for displaying the first candidate character among the candidates on the display section 8; and a display output section for displaying the first candidate character among the characters on the display section 8, and reading out a correction character group corresponding to a predetermined kana input from the kana-kanji conversion dictionary 6 and correcting the correction character group. After being stored in the character storage table 7, the arrangement order of the correction character group in the correction character storage table 7 is changed according to a predetermined rule, and the characters are outputted to the display section 8 in order from the higher order of the correction character group. A control unit 10 including a correction unit is a keyboard device that provides the control unit 8 with predetermined key input such as kana input.

In this embodiment, the character correction section of the control section 8 is
It operates according to the flowchart shown in the figure. Now, as a result of observing a certain input character, as shown in FIGS. 1 and 3(a), a group of candidate characters ranked in the candidate storage memory 5 (indicated as CLAN in FIGS. 2 and 3) It is assumed that Al to AM are stored, and that the first-ranked candidate character A1, which is the recognition result, is different from the input character.

In this case, as shown in Figure 1, a kana input corresponding to the input character is given from the keyboard WIO.
(Fig. 2 ■), the character correction unit reads the correction character groups Bl to 8M and their number corresponding to the kana input from the kana-kanji conversion dictionary 6 (Fig. 2 ■), The correction character groups old to BM are stored in the correction character storage table 7 (indicated as T in FIGS. 2 and 3) in the same arrangement as shown in FIG. 3(b).

Thereafter, the character correction unit refers to the candidate character groups A1 to AN and selects the corrected character storage table? Correction character group 8 in
Change the arrangement order of 1 to BH. At this time, the character correction section includes pointers P1. A pointer P2 indicating the arrangement order of the correction character group Bl to BH before change and a pointer P indicating the advance order for the correction characters are provided, respectively. Pointer P is set to 1, and pointers P1 and P2 are set to O. Each of them is initially set (000 in FIG. 2). In this state, the character correction unit first sets the pointer P1 to 1 and the pointer P2 to 1, and then changes the candidate character CADI(Pi) supported by the pointer p1 to the correction character T( Determine which force) negative force 1 matches P2), and if the two do not match, keep pointer P1 set to 1 and update pointer p2 sequentially from 1 to 1, and then update pointer p1. Pointer P2 is updated sequentially from 1 to 2 while updating from 2 to N, and the above-mentioned judgment operation is repeated (Fig. 2 If the candidate character CAM (PI) supported by P1 matches the correction character T (P2) initially indicated by pointer P2, the character correction section , pointer P
Move the correction letter T (P2) specified in step 2 to the work area W.
After transferring with k (■ in Figure 2), by replacing the above pointer P2 with pointer P3 and updating the pointer P3 in the negative direction, the above correction character string (P2) is placed higher in the correction character storage table 7. The arrangement order of each correction character arranged in i is moved down by -, and the storage area of the correction character storage table 7 corresponding to the arrangement order (X) indicated by pointer P is set to a blank state ( 000・) in Fig. 2, the correction character T(P2) stored in the workpiece gloss Wk is stored as T(P) in the blank storage area (Fig. 2e). .After this,
The character correction section t± updates the pointer P (Fig. 2 @), and thereafter repeats the same judgment operation so that the candidate character CAM (PI) indicated by the pointer PI is
If it matches the correction character T (P2) specified in 2, move the correction character T (P2) to the sequence order specified by pointer P in I/1 in the correction character storage table 7. Deploy.

When such a series of operations is completed, the character correction unit displays the corrected character group of the corrected character storage table lure on the display unit 8 according to the new arrangement order (Fig. 2 @l),
When the displayed correct correction character is selected by key operation on the keyboard device 10, the correction character is redisplayed on the display unit 8 in place of the incorrectly recognized character AI (FIG. 2 O). At this stage, correction of the misrecognized character AI is completed.

Next, the above-mentioned character correction section/process will be explained using a specific example.

Now, it is assumed that the recognition result of the character recognition device for the input form shown in FIG. 4(a) is obtained as shown in FIG. 4(b). In this case, the observed character "hen" is incorrectly recognized as "koi" and the observed character "koto" is incorrectly recognized as "fude", so the operator indicates these misrecognized characters using the keyboard device 10, These characters must be corrected.

First, in correcting the misrecognized character ``koi'', the operator manually inputs ``hen'' as the reading of the correct character ``kin'' using the keyboard device IO. Then, as shown in FIG. 5(b), the character correction unit of the control unit 8 reads out the seven characters [kata, hen, return, odd, bias, hen, hen] from the kana-kanji conversion dictionary 7, and converts these characters. The correction character group is stored in the correction character storage table 7 in the same sequence order. Next, the character correction unit refers to the candidate characters "Koi, Mugi-1 Hen, Fude, Winter" in the candidate storage memory 5 shown in FIG. is moved to the top of the correction character storage table 7, arranged in the order of the correction character group "hen, kata, hen, return, bias, hen, hen" in the correction character storage table 7, and then this The correction character groups are displayed on the display section 8 in order or all at once according to the arrangement order. At this stage, the operator selects the correction character [variant J] that matches the correct character.
When selected by key operation on the keyboard device IO,
The character “weird” is displayed in display section 8 instead of the misrecognized character “koi”.
will be redisplayed, and the text correction will be completed. At this time, since the correction character "kin" that matches the correct character is located at the top of the correction character group, the operator can relatively easily select the correction character.

Next, in correcting the misrecognized character "fude", the operator manually inputs "koto" as the reading of the correct character "koto" using the keyboard device. Then, as shown in FIG. 6(b), the character correction section of the control section 8 reads out the four characters of the collation identification part of "Kan, Koto, Kuruma, Ko" from the Kana-Kanji conversion dictionary 6 and converts them into this arrangement. The characters are stored in the corrected character storage table 7 in their order. Next, the character correction section refers to the candidate characters "Fude, Mine, Higashi, Chu, Kyo" in the candidate storage memory 5 shown in FIG. An attempt is made to change the arrangement, but since the above correction character does not exist in the candidate character group, the above correction character part will change the correction character group according to the original arrangement order without changing the arrangement of the correction character group. are displayed on the display unit 8 sequentially or all at once. At this stage, when the operator selects the correction character ``koto'' that matches the correct character by key operation on the keyboard device IO, the six characters ``koto'' will be displayed again on the display section 8 in place of the incorrectly recognized character ``brush''. will be displayed and the text will be corrected.

Note that in the above embodiment, the reading of a character is used as information for specifying the correct character.

However, there is no problem in using the number of strokes of a character, etc. Further, the specific structure of the character correction section is not limited to that shown in the above embodiment, and it goes without saying that the design can be changed as appropriate.

〔Effect of the invention〕

As explained above, according to the character correction method in the character recognition device according to the present invention, when correcting a misrecognized character, a correct character is selected from among a group of correction characters based on information for specifying the correct character. The ones with a high probability are arranged at the top of the correction character group, and the correction character that matches the correct character is selected from this correction character group.
Not only can you reliably correct characters, but even if
Even when the number of characters in the character group for correction is large, it becomes possible to select the correct character from the character group relatively easily, and the speed of character correction can be improved accordingly.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of a character recognition device for carrying out the character correction method according to the present invention, FIG. 2 is a flowchart showing the operation of the control unit according to the embodiment, and FIG. 3(a) 3(b) and 3(c) are schematic diagrams showing the operation of the control unit for the modified character storage table according to the embodiment, and FIG. 4 is a schematic diagram showing the operation of the control unit for the candidate storage memory according to the embodiment. FIG. 5(a) is an explanatory diagram showing an example of an input form including an observed character and an example of its recognition result. FIG.
) is an explanatory diagram showing an example of the content of the corrected character storage table for the kana reading of the observed character "weird"; FIG. 5(C) is an explanatory diagram showing an example of the content of the corrected character storage table controlled by the control unit; Figures (a), (b) and (C) are explanatory diagrams corresponding to Figures 5 (a), (b) and (c) for the observed character "koto". 1...Observation unit 2...Feature extraction unit 3...Recognition dictionary 4...Verification identification unit 5...Candidate storage memory 6...Kana-Kanji conversion dictionary? ...Corrected character storage table 8...Display section 8...Control section IO...Keyboard device Fig. 3 Fig. 4 Fig. 5 Fig. 61.1

Claims (1)

[Claims] A plurality of candidate characters ranked in descending order of similarity to the observed character are selected from a recognition dictionary, and the first-ranked candidate character from the group of candidate characters is recognized.The resulting character is displayed on the display unit. In the recognition device, when correcting a displayed character that is incorrectly recognized, a group of correction characters including the correct character is selected from a predetermined dictionary based on information for specifying the correct character, and the group of correction characters is selected from a predetermined dictionary. If the above candidate character is included in A character recognition device characterized in that a correction character matching a correct character is selected from among a group of correction characters having How to correct text.