JPH0786908B2 - Word matching device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a device for performing a word matching process based on a reading result of a character string read by a character reading device.

[Conventional technology]

Conventionally, the result obtained by reading a character string on a form by an optical character reading device (hereinafter referred to as OCR) is used for each word by word data in a word dictionary unit in which a word group to be read is registered in advance. There is a word matching device for matching. This type of device uses an index based on the word length of the input word and the character type at a specific position of the input word to search the word dictionary section.

[Problems to be solved by the invention]

By the way, when an erroneous entry is made on the form by mistake, one character may be erased by a vertical line as shown in FIG. 2 (a).
In this case, there is a problem that a character is erroneously collated as one character exists in the erased character portion. In addition, correct characters may be entered in the margin of the erasure character part, and when the writing restrictions on erasure characters are relaxed, it is not possible to distinguish whether the erasure character is considered to have characters or not, and a correct collation cannot be obtained. The problem was that it wasn't.

An object of the present invention is to alleviate the limitation of the writer's correction method using erasure characters while maintaining the reading accuracy of the word matching process in the word matching field.

[Means for solving problems]

According to the present invention, a word dictionary unit in which word data of a word group to be read is registered in advance, and a word matching processing unit that matches the reading result of a character string including a deletion character based on the word data. In the provided word collation device, the word collation processing unit includes a first collating unit that collates based on a reading result of a character string excluding the erasure character and a reject character for which the erasure character is not a candidate character. And a second collating means for collating.

[Action]

According to the present invention, the correct collation can be performed by the first collating means when the erased character is the deletion and by the second collating means when the erased character is the correction.

〔Example〕

FIG. 1 is a block diagram showing an embodiment of the present invention, in which 1 is an input form and 2 is an OC that can be realized by a conventional technique.
R, 3 are word matching processing units for implementing the present invention, 4 is a word dictionary, and 5 is a word matching output. The operation of this method will be described below by taking the reading of an address as an example.

The contents written in the input form 1 are as shown in FIG. 2 (a), and the words are separated by spaces. The reading result of OCR2 at that time is shown in Fig. 2 (b).
The result is as follows. In FIG. 2 (b),
"?" Is a character that cannot be read by OCR2, and "1" is a character that is erased by the writer's correction. In addition, "Ku" of "Totsukaku" indicates a state of being misread by "No".
The contents of FIG. 2 (b) are input to the word matching processing unit 3.

Next, the table contents of the word dictionary 4 are shown in FIG. Here, 6 is a word length index, 7 is a word dictionary, and the word table 7 is sorted and stored according to the word length. In the word table 7, the search start position is indicated by the word length index 6 and the start address where the word of the corresponding word length is stored.

Hereinafter, an example of the operation of the word matching processing unit 3 will be described in detail with reference to the flowchart in FIG.

First, the reading result of the address in FIG. 2 (b) is received by the word collation processing unit 3, the space is regarded as a delimiter between words, and the first word "Yokoha?" At the beginning is extracted to input the word. As W _i , the number of characters other than the erased character L ₁ = 5 and the number of erased character L ₂ = 0. That is, the word corresponds to a word that does not include a deletion character. Based on this, the word length L = 5. The variable FLG in the formula will be described later. Next, the word length index 6 of FIG. 7 is used to obtain the start address of the word having a word length of 5, and the word W _j is read from the word table 7,
Each character of W _i and W _j is compared, and the added value of FIG. 8 is added to the distance value d depending on the state. For example, if the distance value d is calculated between the word "Kamakurashi" in the dictionary and the input word "Yokoha? Shi", there is a mismatch when the first character "ka" and "yo" are compared, and 1 is added according to FIG. Similarly, the second character and the third character are also incremented by one. Since the last character "shi" matches, 0 is added. As a result, the distance value d becomes 3.5. This process is performed for all words having the same word length, and the matched dictionary word and the distance value d are registered as candidate words. When all matching with the word of the same length is completed, and only the registered candidate word having the smallest distance value d exists, the word is output as the correct word. In this case, the distance between the input "Yokohama?" And "Yokohama" in the word dictionary 7 is only 0.5, which is the minimum, and this is used as the result of word matching.

Next, the operation of the present invention will be described using the second and third words "Totsu! Kano" and "Na! First, the word matching is performed like the first word excluding the erasure character. As a result, when there is a word having the only minimum distance value, that word is set as the word matching result, and the only minimum distance value is set. An example will be shown in which, when there is no possessed word, the erased character is not removed and a candidate-free character is used, word matching is performed again, and the result is adopted. For these switching, the derivation formula of the word length L is L = L ₁ + (L ₂
This is realized by controlling the value of FLG in (× FLG) to 0 or 1. Of course, other methods of handling the results of the two types of matching methods can be easily considered.

First, the input word “Totsu! Kano” is an example of rewriting in the next deleted character frame, and the number of characters other than the deleted character L ₁ =
4, the number of erased characters L ₂ = 1 and the word length L = 4 (FLG =
The word matching is performed in 0). In other words, if word matching is performed using the word "Totsukano" excluding "!", The distance value d to the word "Totsukaku" in the dictionary becomes 1 and is the only minimum value, so the same word is used as the result of word matching. .

Next, in the word "Na! Nochiyou" rewritten outside the erased character frame, the number of characters other than the erase character L ₁ = 5, the number of erase character L ₂ = 1, and the word length L = 5 (FLG = 0) But there is a word W _j with only one minimum, the match fails. Therefore, FLG = 1 is set, and further, the word length L = 6 including the correction character by erasure is set and the word matching is performed again. As a result, "Nakanochiyou" is set as the word matching result as the only word with the smallest distance. By the above operation, the content of FIG. 2C is output as the word collation output 5 of the word collation processing unit 3.

〔The invention's effect〕

As described above, since there are two types of correction methods by the erasure character of the writer, there is an advantage that the writing restriction can be relaxed and the accuracy of the word matching process can be improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, and FIG.
4 (a), (b), and (c) are explanatory views showing the input contents of the input form, the contents of the OCR reading result, and the output contents of the word matching processing unit. FIG. FIG. 5 is a flow chart showing the operation of the embodiment, and FIG. 5 is an explanatory diagram showing the added value at the time of word matching. 1 ... Input form, 2 ... OCR, 3 ... Word matching processing unit,
4 ... word dictionary, 5 ... word collation output, 6 ... word length index, 7 ... word table

Claims (1)

[Claims] 1. A word dictionary in which word data of a word group to be read is registered in advance, and a word collation processing unit for collating a reading result of a character string including a deletion character based on the word data. In the word collation device, the word collation processing unit regards the first collation means that performs collation based on the reading result of the character string excluding the erasure character and the erasure character as a rejected character for which no candidate character was obtained. A second collating means for collating the word by means of a second collating means.