JPH0325691A - Correcting method for misread character - Google Patents

Abstract

PURPOSE:To effectively correct the misread characters when many misreading cases occur by storing the misreading contents and the preceding and following character strings to perform the automatic correction with presence of character strings regarded as equal to each other and to urge the corresponding operation with the characters similar to each other. CONSTITUTION:When a misread character is corrected, the misreading contents and the character strings preceding and following the misread character are stored and a character string equal to the misread character is searched out of the subsequent reading results. Then the misread character is automatically corrected if either one of preceding and following characters of the character strings regarded as equal to each other is coincident with the misread character. When both preceding and following characters are not coincident with the misread character, the positions of both characters are shown to decide whether the relevant character is correct or not, whether the correction equal to the automatic correction is required or not, and whether the candidate characters must be selected or not. Then the corresponding operation is urged. Thus the repetitive misreading cases are extracted and corrected automatically. As a result, such kinds of correction can be easily and accurately carried out.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for correcting characters misread by a character recognition device that recognizes and outputs input characters. [Prior art] To correct misread characters after character recognition, an operator indicates the misread character from the recognition result character string displayed on the screen, displays a group of candidate characters for that character, and selects the correct character from among them. The method of selecting is widely used. [Problems to be Solved by the Invention] However, in such a method, even if the same misreading occurs in multiple locations in a document, the operator must discover the misreading location and select candidate characters each time. Overlooks often occur.

Therefore, the problem of the present invention is to extract and correct the same misreadings that occur repeatedly. [Means for solving the problem] When correcting characters that are misread by a character recognition device that recognizes and outputs input characters, first correct the misread characters and memorize the content of the misread and the character strings before and after it. Then, search for the same string as the misread character from the subsequent reading results,
As a result, if one of the characters before or after matches the same character string, it will be automatically corrected, and if the characters before or after do not match, the location will be shown and the character will be correct or automatically corrected. Displays whether a similar correction is required or whether candidate character selection is required and prompts the corresponding operation.

[Effect]

Focusing on the fact that the same misreadings are likely to occur frequently in a single document, the system attempts to efficiently correct misreading characters.It extracts the same misreadings that occur repeatedly and automatically corrects them. This makes this type of correction easy and accurate.

〔Example〕

Fig. 1 is a flowchart showing an embodiment of the present invention, Fig. 2 is an explanatory diagram for specifically explaining the present invention, Fig. 3 is a schematic diagram of a system to which the present invention is applied, and Fig. 4 is its operation. 2 is a schematic flowchart for explaining.

First, let's explain from Figure 3. In the figure, 1 is a scanner that manually inputs a document image into an image memory by optical scanning, 2 is a character recognition device consisting of a recognition unit that extracts and recognizes images of l characters from the document image, and a correction unit that corrects misread characters. , 3 displays the recognition results. 4 is a keyboard for indicating misread characters, etc. The operation is as shown in FIG. 4, in step 1 an image of the entire document is captured, and in step 2 it is divided into images for each character.

Then, in step 3, a comparison is made with a dictionary of character patterns, and as a result, N characters with small inter-feature distances are output.

Furthermore, in step 4, misread characters, which are a feature of the present invention, are corrected, and in step 5, the recognition results are output.

The present invention will be explained below with reference to FIGS. 1 and 2. First, the system corrects the misread character, memorizes the content of the misread character and the character strings before and after it, and then searches for the same character string as the misread character among the subsequent character reading results. Then, if either of the preceding or following characters matches, it will be automatically corrected. FIG. 2 fat shows an example of the reading results of the character recognition device, including misread characters. In this example, "ichi" (long sound) is
(Kanji) has four mispronunciations (see the underlined characters). Three of them appear in the character string "manufacturer". Therefore, the first "one"
(The part marked with ■ in Figure 2 (b)), if this is corrected to the long sound ``ichi'', the ``ichi'' in the character string ``manufacturer'' before and after it becomes 1” automatically. However, some kind of mark, such as shading, will be displayed to indicate that the correction has been made, to prompt the operator to check. In addition, "ichi" in which the preceding and following character strings do not match is "ichi" in "ichilayer".
or “one” in “dealer” (see ■, ■ in Figure 2 (C))
There are two locations like this, but for them, move the cursor to "
As shown in Figure 2(e),
: Correct answer, 2: Correction, 3: Candidate selection instructions are displayed and the operator's judgment is requested. Here, if 1: Correct answer is selected, the answer is left unchanged, and if 2: Correction is selected, correction is made in the same way as automatic correction. Moreover, when 3: Candidate selection is selected, the correct answer can be selected from among the candidate characters as usual.

As in this example, if the misread character is in the middle of a word, the surrounding characters will match, but in reality the misread character is at the beginning of the word. Various cases are possible, such as at the end or at the end of a conjugated word. Therefore, if either one character before or after the same character as the misread character matches the corrected misread character, automatic correction is performed. This is because if it is automatically corrected, it will be displayed with shading, so even if it is wrong, it will be easy to find.

〔Effect of the invention〕

According to the present invention, in the stage of correcting misread characters by the character recognition device, there is an advantage that the more misreadings occur, the more efficiently misread characters can be corrected.

[Brief explanation of drawings]

Fig. 1 is a flowchart showing an embodiment of the present invention, Fig. 2 is an explanatory diagram for specifically explaining the present invention, Fig. 3 is a schematic diagram of a system to which the present invention is applied, and Fig. 4 is its operation. 2 is a schematic flowchart for explaining. Code explanation 1...Scanner, 2...Character recognition device, 3...Display, 4...Keyboard.

Claims (1)

[Claims] 1) When correcting a character that is misread by a character recognition device that recognizes and outputs input characters, first correct the misread character and memorize the content of the misread and the character strings before and after the misread character. search for the same character string as the misread character from the subsequent reading results, and if one of the characters before or after matches the same character string, it will automatically correct it and Misreading characterized by indicating the location of characters that do not match, and prompting the corresponding operation by displaying whether the character is correct, whether correction similar to automatic correction is required, or whether candidate character selection is required. How to correct text.