JPS6145378A - Word reader - Google Patents

Abstract

PURPOSE:To improve both reading accuracy and speed with a word reader by replacing characters of the 1st heading with most similar internal characters if said heading characters are equal to external ones and preparing the 2nd heading having same contents as the 1st heading in other cases within a word dictionary. CONSTITUTION:The characters on a form 1 are scanned by a scanning means 2 and undergo the preprocessing such as the binary coding, etc. through a preprocessing means 3. The recognized candidate characters selected by a character recognizing means 4 are stored to a recognized candidate character memory means 5. A word deciding means 6 decides and delivers a word name out of the recognized candidate characters stored the means 5 by means of a word dictionary 7. The dictionary 7 contains the word name to be read as the 1st heating. In case the external characters are included in those characters which form the 1st heading, these external characters are replaced with the internal characters having similar forms similar to those of the external characters. Thus the 2nd heading having the same contents as the 1st heading is obtained.

Description

[Detailed Description of the Invention] [Technical Field of the Invention] The present invention recognizes characters constituting a word for each character,
The present invention relates to a word reading device that reads words using the results.

[Prior art]

Conventional word reading devices are configured to recognize every seven characters that make up a word, and read vulgar words from the results. Therefore, words containing only recognition target characters (hereinafter referred to as internal characters) can be read, but words containing non-recognition target characters (hereinafter referred to as external characters) cannot be read.In other words, the reading result is It was necessary to correct the reading results to avoid misreading or misdirection.

The easiest way to correct the reading result is to re-enter it with a character code such as JI8 Kuten number. However,
With this method, you have to check the character code you want to re-enter using a comparison table.

The drawback was that it was very time consuming.

In addition, in another method of correcting the reading results, for each internal character, a external character with a similar shape to that character is used in advance, memorized in order of frequency, and when inputting the external character, use the external character with a similar shape to this external character. By selecting and specifying a recognition candidate character, the corresponding external characters are displayed, and the user selects and inputs from among them. This method does not require checking the character code of the external character to be input, but has the disadvantage that it requires human intervention to select a single character.

[Summary of the Invention] The present invention has been made in order to eliminate the above-mentioned drawbacks, and an object of the present invention is to provide a word reading device that can quickly read words including non-standard characters with high accuracy.

The technical means adopted in this invention to achieve this purpose is a word reading device that recognizes and reads words consisting of multiple characters, one by one, each character constituting a word. character recognition means for recognizing and selecting a plurality of recognition candidate characters in descending order of accuracy as input characters; recognition candidate character storage means for storing recognition candidate characters selected by the character recognition means; The target reading word name is the first heading item, and for words that include non-recognition characters among the reading target words, the non-recognition characters are replaced with recognition target characters having similar fonts.

Words that do not include non-recognition target characters include a word dictionary having a second heading item with the same content as the first heading item, a combination of recognition candidate characters in the recognition candidate character storage means, and the combination of the recognition candidate characters in the word dictionary. A word name that matches the second heading item is determined.

This configuration is characterized by comprising a word determining means for outputting the seventh heading item corresponding to the second heading item as a final word reading result.

[Embodiments of the invention]

Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention. The characters on the form are scanned by a scanner stage and subjected to photoelectric conversion, and the preprocessing means 3 performs preprocessing such as binarization, and the recognition candidate characters selected by the two character recognition means are stored in the recognition candidate character storage means. Store in 5. The word determining means 6 uses a word dictionary to determine and output a word name from the recognition candidate characters stored in the recognition candidate character storage means.

FIG. 2 is a diagram for explaining the operation of one character recognition stage ψ. Input word written on form l to be read Zaha 1
For example, the characters ``hema'' t and ``sparrow'' l.

It is assumed that it consists of With character recognition means,
Cut out each character that makes up the input word,
It is recognized using techniques such as already known pattern matching methods, and each character is ranked in descending order of accuracy as an input character (for example, the combination of ``A'' saw and ``Sho'' 13 is shown in the figure). 1) Select the plurality of recognition candidate characters 1) assigned. These recognition candidate characters // are stored in the recognition candidate character storage means 5.

Now, "hemp" 9, which makes up the input word t, is an internal character.

Assuming that "sparrow" IO is a foreign character, "ma" ri is included in the recognition candidate characters lno, but "sparrow" IO is a recognition candidate character l.
It is never included in l.

FIG. 3 shows an example of the structure of the word dictionary 7. This word dictionary 7
has the name of the word to be read as the seventh heading item,
If the characters constituting the first heading item include external characters, replace the external characters with internal characters of similar shape, and create a second heading item with the same content as the other first heading item. It is composed of

Note that the first heading item lq of the word dictionary 7 that includes external characters
To create the second heading item/S corresponding to , the external characters included in the word to be read are recognized.

This is done by replacing the most frequently misread internal characters with external characters in the first heading item.

The word determining means 6 determines a word name in a first order manner using the recognition candidate characters 1) created in one or more ways and the word dictionary 7.

First, among the combinations of recognition candidate characters //, the word dictionary! ! Find a combination that matches the second heading item of Near. Look for matching combinations. If there are a plurality of matching combinations, for example, the combination that minimizes the sum of the ranks assigned to the recognition candidate characters may be selected. In this example, only the combination of the first place "hema" 7.2 and "Ministry"/3 can be found in the second heading item of the word dictionary. In other words, the word ``Mahjong Jtq'' has been determined for the input word t.
By setting the first heading item "Mahjong" III corresponding to the heading item "Mahjong Jts" as the final reading result, the input word za can be read correctly.

In fact, it is known that among the characters that make up a word, the probability of concatenation between kanji characters is less than 7%. There is a very small possibility that the word will be a meaningful word, and therefore the word reading device according to the present invention can read words that include non-standard characters with a sufficiently high probability.

In addition, in the above embodiment, the case of a word consisting of a letter C including one external character was explained, but the present invention is not limited to this, but can also be applied to a word consisting of one letter or three or more letters. This method can also be applied when the characters composing a word include two or more external characters.

〔Effect of the invention〕

As described above, according to the present invention, in the word dictionary.

It has the name of the word to be read as the first heading item, and
If the characters that make up the heading item are external characters, they are replaced with internal characters that are most similar to the character shape, and the other items have a second heading item that has the same content as the first heading item. , it has the advantage of being able to read not only words made up of internal characters only, but also word lines that include external characters with high accuracy and speed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, and FIG. 2 is a diagram for explaining the #h operation of the character recognition means. FIG. 3 is a diagram showing an example of the structure of a word dictionary. In the figure, G... character recognition means, 5... recognition candidate character storage means, 6φ... word determination means, ri... word dictionary f a group input word /10m recognition candidate characters. They are shown with the same reference numerals. Procedural amendment (voluntary) Showa 6゜chi4, May 5th

Claims (2)

[Claims] (1) A word reading device that recognizes and reads words made up of multiple characters, which recognizes each character that makes up a word and selects multiple recognition candidate characters in order of their accuracy as input characters. a recognition candidate character storage means for storing recognition candidate characters selected by the character recognition means; and a recognition candidate character storage means for storing recognition candidate characters selected by the character recognition means; For words that include non-recognized characters, the non-recognized characters are replaced with recognition target characters that have similar glyphs, and words that do not include non-recognized characters are created as second heading items with the same content as the first heading item. a combination of recognition candidate characters in the recognition candidate character storage means and the second word dictionary in the word dictionary;
word determining means for comparing the first heading item with the first heading item to determine a matching word name, and outputting the first heading item corresponding to the second heading item as a final word reading result. A word reading device characterized by: (2) The word determining means compares the combination of the recognition candidate characters and the second heading item, and if there are multiple combinations in which both match, the word determination means determines the rank assigned to the recognition candidate characters. The word reading device according to claim 1, which selects a combination with a minimum sum.