JPS59106083A - Word reader - Google Patents

Abstract

PURPOSE:To read easily a word containing an external character, as well, by deriving a similarity degree of a word unit by use of a similarity degree of an internal character in a similar character table, in case of a word containing an external character. CONSTITUTION:A word recorded on a business form 1 is scanned 2, converted to a character pattern of every one character, and sent to a similarity degree calculating device 3. This device 3 calculates a similarity degree of every one character from an input character pattern 12 and a reference pattern in a recognition dictionary 4, and sends only a character having a large similarity degree as a proposed character 13 together with the similarity degree to a word determining circuit 5. As a result, the circuit 5 refers to the contents of a word dictionary 6, and determines a similarity degree of a word unit. Also, in case when an external character is contained in a word in the word dictionary 6, a similarity degree of the word unit is derived by investigating the contents of a similar character table 7, checking an internal character having a character shape similar to its external character, and using a similarity degree of this internal character. In this way, a word whose similarity degree becomes the maximum becomes a read result.

Description

[Detailed Description of the Invention] [Field to which the invention pertains] The present invention recognizes each character constituting a word,
The present invention relates to a word reading device that reads words from the results.

[Configuration and operation of conventional technology]

Conventionally known word reading devices of this type are configured to use a recognition dictionary to recognize each character constituting a word, and to read the word based on the results. And, in the word reading device mentioned above,
Words consisting only of characters to be recognized (hereinafter referred to as internal characters) included in the recognition dictionary for character recognition can be read, but characters other than the characters to be recognized that are not included in the recognition dictionary (hereinafter referred to as external characters) are readable. Words containing `` are constructed so that they cannot be read except in special cases.

For example, when reading the two words ``consumption'' and ``extinction'', even though the character ``expense'' cannot be recognized because it is an external character, each character ``expense'' and ``extinction'' are internal characters. If recognition is possible, it is possible to read the above two words. However, if the character ``expense'' or the character ``mei'' is an external character, and the character ``era'' is an internal character, it is impossible to read the two words. This actually means that each character "erases"
and ``expense'' are included in the educational kanji, but the character ``metsu'' is not included, so a word reading device with a recognition dictionary that only contains about 1000 characters as internal characters will be able to read the above two words. That becomes impossible. Of course, increasing the number of internal characters can make reading possible, but in general, in order to increase the number of internal characters by one character, it is necessary to increase the storage capacity of the recognition dictionary from several tens of bytes to several hundred bytes. there were.

[Disadvantages of conventional technology]

Since the conventional word reading device is configured as described above, as mentioned above, it is generally only able to recognize words consisting of internal characters, and it is also possible to increase the number of internal characters. This inevitably requires an increase in the storage capacity of the recognition dictionary, which has the disadvantage of making the device larger and more expensive.

[Object of the present invention]

The present invention has been made in order to eliminate the drawbacks of the conventional ones as described above. The apparatus includes a similar character table that defines internal characters with similar characters and shadows to external characters, and a similarity calculation means for calculating the similarity of each character from an input character pattern and a reference pattern in the recognition dictionary. For words that include external characters, we break down the structure in which the similarity of internal characters in the similar character table is used to find the similarity of each word, and make it possible to easily read words that include external characters. DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a word reading device which is an embodiment of the present invention. In Figure 1, 1 is a form, 2 is a form 1
a scanning means for scanning the words recorded above; 4 a recognition dictionary storing standard patterns of internal characters; 3 calculating similarity for each character from the input character pattern and the standard pattern L in the recognition dictionary 4; Similarity calculation means, 6 is a word dictionary storing words including non-standard characters, 7 is a similar character table that defines internal characters similar in shape to the non-standard characters, 5 refers to the word dictionary 6, This is a word determining means that determines the similarity of each word.

FIG. 2 is a diagram showing an example of part of the contents of a word dictionary in the word reading device of FIG. 1. Word dictionary 6 shown in Figure 2
is expressed by a combination of character codes 8 for each character. In this example, the character code 8 is J I 8
The kuten number specified in C6229 is attached, and the above J
According to ISC6229, the Kuten band code of 1601 to 8394t is defined as the character code of Kanji. The symbol n in FIG. 2 is a non-standard character number 9 indicating that it is a non-standard character, and the relative value is 8401 or more. As a result, if it is 8401 or more, it is a custom character, and 8394
If the following is true, it can be determined that the character is an internal character. Here, the character "Mei" is used as an external character, and the other characters "Ku" and "Kai" are used as internal characters.

FIG. 3 is a diagram showing an example of a partial content of the similar character table in the word reading device of FIG. 1.

The similar character table 7 shown in FIG. 3 includes the above-mentioned external character number 9. Custom character code 1o which is the character code of the relevant custom character
and a similar internal character code 11 which is a character code of an internal character similar in shape to the external character. If the character code in the word dictionary 6 is 8401 or higher, this is determined to be a non-standard character number 9, and the character with the similar internal character code 11 set in the position of the corresponding non-standard character number 9 in the similar character table 7 is using the similarity measure.

Fig. 4 shows the similarity calculation means in the word reading device shown in Fig. 1, which shows candidate characters arranged in the order of the values obtained by calculating the similarity between the Kugata character pattern and each character of the internal characters. FIG. 4, which is a diagram showing an example, shows candidates in which the degree of similarity between the character pattern 12 of the character "Mei" and each character in the internal characters is calculated, and the candidates are arranged in order from the one with the largest value. As character 13, 6 from the character "reduced" to the character "shallow"
characters are shown. The numerical value in parentheses below the candidate character 13 in FIG. 4 indicates the degree of similarity 14 with each candidate character 13. The operation of the word reading device according to the embodiment will be explained using the above-mentioned FIGS. 1 to 4. First, the words recorded on the form 1 shown in FIG. This similarity calculation means 3 calculates the similarity for each character from the input character pattern 12 shown in FIG. It is sent to the word determining means 5 together with the similarity. Here, as a matter of course, the degree of similarity is calculated only for internal characters, so the characters sent to the word determining means 5 are only internal characters. The word determining means 5 refers to the contents of the word dictionary 6 shown in Fig. 2, and determines the degree of similarity of each word. This degree of similarity is assumed to be, for example, the average value of the degrees of similarity for each character. If a word in the word dictionary 6 includes a non-standard character, refer to the contents of the similar character table 7 shown in FIG. Find the similarity of each word using the similarity. Then, the word with the highest degree of similarity is taken as the reading result.

Therefore, for example, when reading the word "extinction" as an input character, if the similarity of the inner character "extinction" is 0.9, the average similarity to the word "extinction" is shown in Figure 4. Using the similarity of 0.8 for the similar internal character ``Ku'' instead of the external character ``Mei'', we obtain (0,9+
0.8)/2=0.85.

In addition, the average similarity to the word "consumption" is (0,9+0.55)/2 # 0, since the similarity of the external character "mei" to the character "cost" is less than 0.55.
．． It will be less than 73. From this result, it is possible to correctly read the word "extinction" that includes external characters because the average similarity becomes large as described above.

By the way, the above-mentioned similar character table 7, which is a main component of the present invention, calculates the degree of similarity between the external character pattern of the input character and the reference pattern 7 of internal characters stored in the recognition dictionary 4 by using the similarity calculation means 3. It can be easily created by calculating the similarity value and selecting the character whose calculated similarity value is the largest on average. Furthermore, since the similar character table 7 is specifically stored in a storage device, the table 7 in FIGS.
The external character number 9 shown in the figure is converted into a storage device address by performing appropriate arithmetic operations. As a result, the storage capacity of the similar character table 7 is 4 bytes for one external character, compared to the tens to hundreds of bytes of 8-byte storage capacity when storing the reference pattern as an internal character in the recognition dictionary 4. This makes it possible to significantly reduce storage capacity.

[Other embodiments of the present invention]

In the above embodiment, the case where the word-based similarity is the average similarity of each character is described, but the present invention is not limited to this, and the word-based similarity is Similarity may also be defined.

In addition, although the above embodiment describes the case where the input characters are two kanji characters, it can also be applied to characters other than kanji or words consisting of three or more characters, and in this case, the same method as in the above embodiment can be applied. It has the effect of

[Effects of the present invention]

As described above, according to the word reading device according to the present invention,
For words that include external characters that are not included in the recognition dictionary, a similar character table is provided that defines internal characters that are similar in shape to the external characters, and the similarity of the internal characters is used to calculate the similarity of each word. Since the structure is made up of . In some cases, the word reading device of the present invention can achieve remarkable high performance when applied to reading words that are composed of character types that have a large number of character types compared to the frequency of occurrence, such as addresses and two names. It has excellent effects.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a word reading device which is an embodiment of the present invention, FIG. 2 is a diagram showing an example of partial contents of a word dictionary in the word reading device of FIG. 1, and FIG. FIG. 4 is a diagram showing an example of a partial content of a similar character table in the word reading device shown in FIG. 4. FIG. 4 is a similarity calculation means in the word reading device shown in FIG. FIG. 3 is a diagram illustrating an example of candidate characters arranged in the order of values obtained by calculating similarity. 1......Form, 2...Scanning means, 3-...Similarity calculation means, 4......
・・Recognition dictionary, 5・・・・・・・・Word determination means, 6
・・・・・・・・・Word dictionary, 7・・・・・・Similar character table, 8・・・・・・Character code, 9
・・・・・・・・・−・Gaiji character number, 10・・・・・・・・・・
・External character code, 11...Similar internal character code, 12...Character pattern, 13-
......- Candidate character, 14 ...... Similarity value. In addition, in the figures, the same reference numerals indicate the same or equivalent parts. Agent Makoto Kuzuno - Figure 1 67 Figure 2

Claims (1)

[Claims] In a word reading device that recognizes and reads words printed on paper, etc., there is a recognition dictionary that stores reference patterns of characters to be recognized, a word dictionary that stores words that include characters other than the characters to be recognized, and a recognition dictionary that stores a reference pattern of characters to be recognized. For characters other than letters, similarity is calculated for each character based on a similar character table that defines recognition target characters with similar glyph shapes to the character, and an input character pattern and a reference pattern in the recognition dictionary. When W talc a word by calculating the similarity of each word from the calculated similarity of each character, for words containing characters other than the recognition target characters, A word reading device characterized in that a word similarity is calculated using a similarity of characters to be recognized.