JPH0632092B2 - Word matching method - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a word matching method, and particularly when there are a plurality of candidate characters for the character position of an input character string, even if there is an error in the input character, The present invention relates to a word recognition method for efficiently recognizing words.

[Conventional technology]

As a conventional word recognition method, when recognizing a word from an input character string represented by a sequence of character codes, there are a plurality of candidate characters for each character position of the input character string, and a typographical error in the input character string, In some cases, it is necessary to collate each candidate character at each position of the input character string with the characters at one or more positions that form an arbitrary word, assuming that the characters include omitted characters and surplus characters.

FIG. 6 is a block diagram showing an example of such a conventional word recognition device.

In FIG. 6, 1 is a word selection table, 2 is a selection processing unit, 3 is a judgment value table, 4 is a judgment value accumulation processing unit, 5 is a sort processing unit, and 6 is an evaluation. It is a value table. For example, if "Decotal" is input as the input character string (it is assumed that the character "ko" is mistakenly input as "ko"), the word selection table 1 of the selection processing unit 2 is referred to and the first character is input. Second character ...
.. outputs the word code and the judgment value to each of them, and transfers them to the judgment value accumulation processing unit 4 in the next stage. The judgment value accumulation processing unit 4 determines the evaluation value for each input word code by referring to the judgment value table 3, and transfers the evaluation value to the next-stage sort processing unit 5. The sort processing unit 5 ranks and outputs the candidate words according to the evaluation value table. The highest ranked are the recognized letters.

In this word recognition device, in order to compare each character code sequence of the input character string and each word using a word dictionary in which each character code sequence is associated with each word, each character position of the input character string Enter the character code that constitutes the input character string in. That is, as described above, when "Decotal" is input, the digital W11, the data W12, and the decoder W13 output for the first character "De", and the decoder W13 for the second character "Co". W13 and W25 of the record are output. Thereafter, in the same manner, the word code and the judgment value are output for each input character. in this way,
The selection processing unit 2 outputs the word code including the characters of the character code at each character position by using the word selection table 1.
In this word recognition device, since it corresponds to erroneous characters, surplus characters, and missing characters, the character position in the word code in the dictionary (hereinafter, referred to as the position in the dictionary) and the character position in the input character string (hereinafter, referred to as the position at the time of recognition) ) Does not match, the word selection table is searched at one or more positions in the dictionary for each candidate character at the recognition position.

For the conventional word matching method, refer to, for example, the specification of Japanese Patent Application No. 61-248415 filed previously and the drawings (title of the invention: word recognition device).

[Problems to be Solved by the Invention]

In such a word recognition device, there is at least one candidate character at the recognition-time position, and the word selection table is searched at one or more positions within the dictionary for the candidate character. An incorrect evaluation value will be output as described below. Therefore, there arises a problem that such an erroneous output should be corrected.

FIG. 7 is a diagram showing a specific processing procedure by the conventional word recognition device.

In character recognition, as shown in FIG. 7, one or more character candidates are output at one recognition position. That is, if there are Tokyo, Kyoto, and Tokyo in the word dictionary and "Kyoto" is entered as the input character string, the candidate characters in the recognition-time position c1 (the first character) are three characters, east, Kyo, and Toru. , C2, three characters, that is, capital, county, and group, are output. In other words, there is a case where only candidate characters within the same recognition position are satisfied as a word (“Tokyo”). In this case, when the above method is used, in the selection processing unit 2 of FIG. 7, when the first character at the recognition position c is regarded as the first character in the dictionary, “Tokyo” is extracted from the candidate characters “east”, When c1 is regarded as the second character in the dictionary, “Tokyo” is extracted from the candidate characters “Kyo”. At this time, although only one candidate character can be used at one recognition-time position, two candidate characters are used at one recognition-time position in the search for "Tokyo", resulting in incorrect evaluation. A value will be given. That is, in the judgment value accumulating unit, "Tokyo" is evaluated as 1.7, "Kyoto" is evaluated as 1.6, and "Toshin" is evaluated as 0.8, and in the next sort processing unit, "Tokyo" is evaluated as 1.7,
"Kyoto" will be 1.6 and "Toshin" will be 0.8. Therefore, the conventional method does not perform accurate processing.

Due to this, in the example of FIG. 7, “Kyoto” which should be originally extracted as the first place is output as the second place.

An object of the present invention is to solve such a conventional problem and to perform collation of characters at one or more in-dictionary positions with respect to one recognized position without incorrectly outputting an evaluation value and It is to provide a word matching method capable of recognizing a word with high accuracy even if there is a typographical error, omission, or surplus character, and also capable of recognizing a word with high accuracy even when performing parallel processing. .

[Means for Solving the Problems]

In order to achieve the above object, the word matching method of the present invention is a word matching method in which an input character string in which a character code is arranged is matched with which word among preset words. It includes a word selection table that associates the used characters with words for each position of the character array that is formed (hereinafter, referred to as a position in the dictionary), and any character position of the input character string (hereinafter, referred to as a recognition position). ) Inputting the character code of the candidate character and the position in the dictionary, the selection processing means for outputting the corresponding word code, and the position where the recognition time position of the character code used for selecting the word code is recorded for each word code. Including a history table, the word code and the position at the time of recognition of the character code used to select the word code are input,
For the position at the time of recognition of the word code, a position verification means for outputting a search history indicating whether or not another candidate character has been searched for, and an evaluation value indicating the certainty for each word code output from the position verification means are displayed. It has a judgment value accumulating means for adding, and a sorting means for extracting a candidate word based on the magnitude relationship of the judgment values of the judgment value accumulating means, and for the candidate character for each recognition position of the input character string by the selection processing means, A word code corresponding to a character code is selected from the word selection table for each position in the dictionary at one or more locations, and the position history detecting means reads the search history of the word code and the recognition-time position from the position history table. Only when it is not searched, the searched history is set in the position history table, and the word code is sent to the judgment value accumulating means. The evaluation value of the word codes output from test unit and accumulating obtain a determination value for the input character string, is characterized in that extracting the candidate words by the sorting means. Further, the search history of the position history table is characterized by writing a measure of certainty obtained from the recognition result of each character of the input character string.

[Work]

In the present invention, the accuracy rate is improved by introducing a method that allows only one search for one recognition-time position. That is, a position history table that holds the score and the recognition position for each candidate character for each recognition position of the input character string and records the recognition position of the character code used to select that word code for each word code. , And input the word code and the position at the time of recognition of the candidate character used to select that word code, and regarding that word code,
The word code whose position at the time of recognition has not been searched, and the position verification unit that outputs the evaluation value of that word code, the word code obtained by the selection processing unit, and the position at the time of recognition of the candidate character that selected that word code. From that, it is verified whether selection from the same recognition position as the recognition position has already been made, and only when the selection is made for the first time, the word code and evaluation value are output to the judgment value accumulating unit, and at the same time. Record verified in the position verification table.

As a result, there is at least one candidate character recognition at each recognition position of the input character string, and typographical errors, omissions,
Even if the surplus characters are included, the erroneous evaluation of the word code that occurs when one or more candidate characters at one recognition position and the character at one or more positions in the dictionary are compared can be stored in the position history table. By avoiding using it, highly accurate word recognition becomes possible.

〔Example〕

 Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block configuration diagram of a word matching device showing an embodiment of the present invention.

The word matching device of the present invention includes a selection processing unit 2 including a word selection table 1 and a position verification unit 7 including a position history table 8.
And a judgment value accumulating unit 4 including the evaluation value table 3 and a sort processing unit 5 including a candidate word table (not shown).

FIG. 2 is a process chart showing an embodiment of the word matching system of the present invention.

Now, the operation principle of the present invention will be described for the case where the same character recognition result is obtained for the same input character as in FIG.

The selection processing unit 2 performs the following selection processing for each recognition-time position. First, the word selection table 1 is searched by the character code of each candidate character at the recognition position, and the word code corresponding to the character code is read. The word selection table 1 shows a correspondence table of a character code for each position in the dictionary and a word code in which the character code exists at the position in the dictionary.
The word recognition device selects a word for the recognition-time position c1 when the position in the dictionary is the first character and the second character. Then, the candidate characters "East" to "Tokyo" and the candidate characters "Kyo" to "Kyoto" are read out respectively, and "Tokyo", the evaluation value 0.9, and the recognition position c.
1, 'Kyoto', evaluation value 0.8, and c1 are output in combination. In addition, in, “Tokyo” is read from the candidate characters “Kyo”, and “Tokyo”, 0.8, and the recognition time position c1 are combined and output. Also, the recognition position c2
For, the word selection table 1 having the first and second characters in the dictionary outputs "Tokyo", 0.8 and c2, or "Kyoto", 0.8 and c2 in combination. In this way, by sequentially processing, the selected word code, the position at the time of recognition, and the evaluation value are paired,
It is transferred to the position verification unit 7 in the next stage.

The position verification unit 7 retrieves the search history relating to the recognition time position of the word code from the position history table 8 by the word code sent from the word selection unit 2, and uses the recognition time position sent from the word selection table 1. Tests if the search for was previously done. The position history table 8 shown in FIG. 2 is an example in which the position at the time of recognition is up to the second character, and for each word code, the processed positions (1)
It has a field that can record whether it is unprocessed (0).
As an example of the position verification process using the position history table 8, the position history table 8 is in a state where the search history obtained by performing the position verification is recorded as shown in FIG. The case where the code is "Tokyo", the evaluation value is 0.8, and the recognition position is c1 is sent. The position history table 8 determines that the word “Tokyo” is duplicated from the word code “Tokyo”, and the recognition time position c1 has been processed (1).
It is obtained that 2 is unprocessed (0). From this,
In the word code “Tokyo”, it can be detected that the recognition-time position c1 sent from the selection processing unit 2 has been processed, so the word code “Tokyo” is not sent to the determination value accumulating unit 4. Further, in the example in which the position test is performed, for the word code'Kyoto ', it is detected that the recognition time position c2 sent from the selection processing unit 2 is unprocessed, and the word code'Kyoto' is evaluated together with the evaluation value. It is sent to the judgment value accumulating unit 4.
At this time, searched (1) is set in the column designated by “Kyoto” and c2 in the position history table 8.

The judgment value accumulating unit 4 cumulatively adds the evaluation values for the same word code in the word data output by the position verification unit 7. In the above example, the word code "Tokyo" was sent only in the case of Fig. 2, so the evaluation value was 0.9, and the correct value was given as the evaluation value of the word obtained from the recognition candidate characters. become. After that, the sort processing unit 5 compares the evaluation values. In the example of FIG. 2, this word recognition device outputs “Kyoto” as the first candidate.

FIG. 3 is a diagram of a position history table at the time of expansion in the present invention.

That is, FIG. 3 shows the position history table when the recognized position is expanded to the Nth character. The position history table 8 is composed of columns for recording processed (1) or unprocessed (0) for each word code W and each arbitrary recognition-time position C. When searching the position history table 8, 1/0 can be obtained by using the word code and the currently-recognized position at the time of recognition. Therefore, when the word code Wk is selected by the selection processing unit 2 by inputting the character code CDj of the candidate character at the recognition-time position Ci, the value in the Wk row and Ci column of the position history table 8 is read out. , It is verified whether or not the word code is output to the judgment value accumulating unit 4. If it is unprocessed as a result of the test, the search completed (1) is recorded in the Wk row and the Ci column, and it is confirmed that all the recognition positions C1 to Cn have been processed, and the word code and evaluation are performed. The value is sent to the judgment value accumulating unit 4. If the input word code has already been processed from the table, the word code is not output to the judgment value accumulating unit 4.

The configuration of FIG. 1 will be described in more detail.

In the configuration of FIG. 1, each candidate character at each recognition position of the input character string, a recognition position, and a set of evaluation values are input, and the most certain word is output. For each candidate character code of each recognition position, the search circuit 11 uses the word selection table 1 in the memory to search for a word code corresponding to the character code for one or more positions in the dictionary, and to find the searched word code. It is sent to the judgment circuit 12. The judgment circuit 12 refers to the position history table 8 in the memory and verifies whether or not the word code is duplicately searched. When the duplicate search is not performed, the search result is recorded in the position history table 8 and the word code and the evaluation value are sent to the adding circuit 13. The adder circuit 13 detects the judgment value up to the present of the word code from the evaluation value table 3 in the memory, updates the word code judgment value by adding the evaluation value, and returns the value to the evaluation value table 3. . After the processing of all the positions at the time of recognition in all the dictionaries is completed, the sorting circuit 9 sorts the determination value table 6 (not shown) and outputs the result. That is, the word code and the determination value are output.

FIG. 4 is a block configuration diagram of a word matching device showing another embodiment of the present invention, and FIG. 5 is a diagram of a position history table used in FIG.

FIG. 4 shows a configuration example in the case of collating in parallel.

In FIG. 4, the selection processing / position verification units 10a-10
n search circuits 11a-11n and determination circuits 12a-1
2n is used and executed in parallel for each in-dictionary position to create position history tables 8a to 8n for each in-dictionary position.
After that, the adder circuit 13 calculates a judgment value, the sorting circuit 9 compares the judgment values, and outputs the result. At this time, the position history tables 8 and 8a to 8n
As shown in FIGS. 5A, 5B, and 5C, the evaluation value is recorded in the column of each recognition time position of the word code, instead of the 1/0 flag.

After the processing for each position in each dictionary is completed, the position history tables 8a to 8n are inspected in the comparison circuit 14, and when the priority is given, the evaluation values are compared, and the high evaluation value is prioritized. Avoid false evaluation due to duplicate search. As a result, it becomes possible to perform collation with a candidate character having a high score at the time of recognition.

In FIG. 5, the word code W of the first character in the dictionary position
The first character at the recognition position of 1 and the first character at the recognition position of the word code W1 of the Mth character in the dictionary overlap. Therefore, comparing both, 0.8 with a high score is the position 1 in the dictionary.
The score of the character is the score of the W1 row and the 1st column.

It should be noted that character recognition at one or more recognition positions with one or more positions in the dictionary occurs when recognition occurs because character or syllable punctuation is unknown, such as handwritten characters at arbitrary pitches or continuous speech. It is also necessary when correcting the deviation between the position and the position in the dictionary.

Thus, in the present invention, by providing a position history table recording the search history of the recognition time position for each word code, by inputting the recognition time position of the word code and the character code selected word code, Whether or not there is a duplicate search is tested using a position test unit that outputs the unsearched word code and the evaluation value for the recognized position of the word code. As a result, when there is a plurality of candidate characters in one recognition position and there is a word formed by combining the candidate characters, when collating characters at one or more dictionary positions for one recognition position. Further, even when the evaluation value is erroneously output, the position detection unit refers to the search history from the position history table based on the word code and the position at the time of recognition, so that erroneous evaluation can be prevented. Even when performing parallel processing, each processor has its own position history table, and the value recorded in the position history table is used as the evaluation value of the character candidate at each recognition position. Can recognize words.

〔The invention's effect〕

As described above, according to the present invention, even if there is a character candidate of one code or more at each character position of the input character string, and there is a typographical error, a missing character, or a surplus character in the input character string, the word can be accurately generated. Can be recognized. Further, even in the case of performing the parallel processing, the word can be recognized with high accuracy by using the value recorded in the position history table as the evaluation value of the character candidate.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block configuration diagram of a word collating device showing an embodiment of the present invention, FIG. 2 is a process diagram when a recognition process is performed using FIG. 1, and FIG. 1 is a diagram showing the contents of the position history table in FIG. 1, FIG. 4 is a block diagram of a word collating device for parallel processing showing another embodiment of the present invention, and FIG. 5 is a position history used in the device of FIG. FIG. 6 is a block diagram of a conventional word recognition device, and FIG. 7 is a diagram showing a table configuration example.
It is a figure which shows the operation process of the apparatus of the figure. 1: word selection table, 2: selection processing unit, 3: determination value table, 4: determination value accumulation unit, 5: sorting processing unit, 6: position history table, 7: position verification unit, 8, 8a to 8n: position History table, 9: sort circuit, 10a to 10m: selection processing / position verification unit, 11: search circuit, 12: determination circuit, 13: addition circuit, 14: comparison circuit.

Claims (2)

[Claims] 1. A word collation method for collating which word of preset words an input character string in which character codes are arranged corresponds to each position (hereinafter, Each word in the dictionary) includes a word selection table in which the used characters correspond to words, and the character code of the candidate character at an arbitrary character position in the input character string (hereinafter, referred to as the recognition position) and the dictionary The word code and the word code include a selection processing means for inputting the position and outputting a corresponding word code, and a position history table recording the recognition time position of the character code used for selecting the word for each word code. The position verification means for inputting the recognition time position of the character code used for the selection, and outputting the search history of whether the word code recognition time has been searched with other candidate characters, and the position verification means. For each word code output from the column, a judgment value accumulating means for adding an evaluation value indicating the certainty, and a sorting means for extracting candidate words based on the magnitude relation of the judgment values of the judgment value accumulating means, For the candidate character for each recognition position of the input character string by the selection processing means, 1
A word code corresponding to a character code is selected from the word selection table for each position in the dictionary at more than one place, and the position history detecting unit reads the search history of the word code and the recognition-time position from the position history table. Only when it is a search, the searched history is set in the position history table, the word code is sent to the judgment value accumulating means, and in the judgment value accumulating means, the word code output from the position verification means. Is added cumulatively to obtain a judgment value for the input character string, and a candidate word is extracted by the sorting means. 2. As a search history of the position history table,
Claim 1 characterized by writing a measure of certainty obtained from the recognition result of each character of the input character string.
The word matching method described in the section.