JPH0193876A - Character reader - Google Patents

Abstract

PURPOSE:To improve the accuracy of a word collating result by adding plural similar characters to a recognizing candidate obtained by a character reading means to the recognizing candidate. CONSTITUTION:When the plural recognizing candidate characters are outputted from a character recognizing means 2, a candidate selecting means 4 defines the recognizing candidate character to be a key word and indexes a similar character table 5. The characters registered in the similar character table 5 are considered to be the candidate characters and a word collating means 6 collates the candidate character to a word dictionary 7 to output the most matching word as a collated result word 8.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a character reading device that recognizes and reads characters written on a form, and particularly to a character reading device that can accurately recognize input characters using a word dictionary. The present invention relates to a character reading device equipped with such a character recognition device.

[Conventional technology]

In conventional character recognition methods, in order to improve the accuracy of the results read by a single character reading device, as described in Japanese Patent Publication No. 61-20058, multiple candidates are output and compared with a word dictionary to find the best match. It was outputting words of remembrance.

[Problem that the invention seeks to solve]

However, in the above method, it is necessary that the recognition candidates include characters of the correct category, and if there is not, there is a high possibility that the characters will be read incorrectly.

- There is a limit to the number of character types that can be read by character reading devices, and only common kanji or JIS level 1 kanji can be read. However, the characters written on the form may include other characters11, such as JIS level 2 kanji.

Restricting the writing of these characters would place an unnecessary burden on the person writing them, and remembering which characters can be read by a character reading device would be difficult since there are thousands of character types. It's almost impossible.

SUMMARY OF THE INVENTION An object of the present invention is to provide a means for performing word matching without erroneously recognizing characters even if the correct characters are not included in candidate characters for recognition from a character reading device.

[Means for solving problems]

The above object is to provide a character reading device with candidate selection means for adding a plurality of similar characters similar to the recognition candidates obtained by the character reading means to the recognition candidates,
This is accomplished by selecting the word with the most similarity from among the words stored in the word dictionary based on the t candidate groups created by the method.

[Effect]

In general, characters in one category are mistakenly read as other categories, but the number of other categories is very large (and often limited to a specific category. Therefore, it is necessary to measure these in advance and read similar characters. When a table is created and R is run, the probability that a correct character will appear in the similar character table becomes very large.

Similarly, characters that are not registered in standard buttons (external characters)
To find out which characters are mispronounced by having them read it, put the foreign characters in the similar character table. (Kotoka De Saru.

If multiple recognition candidate characters are output from the character reading device, the recognition candidate characters are used as keywords to retrieve all similar characters, and the characters registered in these similar character tables are also used as candidate characters. Deciding to do so,
Correct characters will be included in the complement characters with a high probability, and there will be no mistakes in word matching.

[Fu example]

An embodiment of the present invention will be described below with reference to FIG.

In Figure 1, 1 is a neatly written ledger.

It is captured and read by the character recognition means 2. The character recognition means 2 performs recognition using a recognition dictionary 5 and outputs recognition candidates. It is assumed that the recognition candidates are composed of a category name, a mask number, a degree of similarity, and the like. The mask number is used to indicate which mask is used when the specimen button is composed of multiple masks per category, and is unnecessary when only one mask is used for the category. , 6, the word matching means matches the candidate characters with the word dictionary 7, and the word that matches best is the matching result word 8.
This is what is output as.

The candidate selection means 4 subtracts the similar character table 5 from the recognition candidates from the character recognition means 2 and creates a candidate character group to be input to the word matching means 6.

FIG. 2 shows an example of recognition candidates from the character recognition means 2. The recognition candidates consist of the number of candidates NC and N, C category names, mask numbers, and similar skins.

FIG. 6 shows an example of a similar character table 501.

The similar character table is composed of a category name, a mask number, the number of similar characters HB, and N5 similar character category names.

The specific operation will be explained below using an example. Suppose that when we read the word ``city'' written on the form, we get the candidates shown in Figure 2. Since the correct answer ``city'' is not included in these candidates, there is a possibility that the word matching may be mistakenly matched with another word.

Next, the first character in the recognition candidate, category name 'Cloth', mask number 1' (!-) is searched from the similar character table 5. It is listed in the similar character table shown in FIG. Aji, four similar characters, and lower grade repetition are found.Among these similar characters, emperor exists among the recognition candidates, so it is added to the remaining lower grade pension candidate table.

The final candidate table is shown in Figure 4. Does this include the correct category ``city''? There was a mistake in word matching, and Maroko was nothing.

In the above example, we took the case where the correct character is not included in the candidates, but even if the written character is a non-standard character,
It is clear that if you include external characters in the similar character table, you can use them in the same way.

(Effects of the Invention) According to the present invention, since it is possible to infer from recognition candidates and add them as similar character-sound candidates, even if the correct character is not included in the recognition candidates, there is a high probability that the correct character will be among the candidate characters. By including non-standard characters in the similar character table, word matching can be performed successfully even if non-standard characters are written. Can be done.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a simple embodiment of the present invention. Figure 2 is an explanatory diagram showing all examples of recognition candidates, Figure 5 is an explanatory diagram showing an example of a similar character table, and Figure 4 is an example of a group of nine candidates created from recognition candidates and similar character table summaries. FIG. 1...Ledger. 2...Character recognition means. 4...Candidate selection means. 6...Word matching means. Deputy Rinshi Katsuo Koyo 2A Departure: 5 Inokumaku 4 En

Claims (1)

[Claims] 1. A character reading means for recognizing characters written on a form;
In a character reading device equipped with a matching means for detecting that a character read by the reading means matches a word registered in a word dictionary, a plurality of similar recognition candidates are selected from recognition candidates obtained by the character reading means. It is characterized by having a candidate selection means for adding similar characters to the recognition candidates, and selecting the word with the greatest similarity from among the words held in the word dictionary based on the candidate group thus created. Character reading device.