JPH02129756A - Word collating device - Google Patents

Abstract

PURPOSE:To prevent a synonym dictionary from increasing its storage capacity and to execute word collation considering the meaning of a word by dividing the word to be collated into nouns consisting of short units, replacing a part or the whole of the word by its synonym and then collating the character string. CONSTITUTION:A word A is divided into nouns consisting of short units by a divided writing part 100. A synonym replacing part 200 searches the synonym of the noun of the short unit by using its internal synonym dictionary. Preferably, a long unit noun coupling plural short unit nouns is formed and the synonym of the long unit noun is obtd. from a synonym dictionary. The work formed by the synonym replacing part 200 and including a synonym in a part or the whole of the word and the original word A are sent to a decision part 300 and compared with a word B. Thereby, it is unnecessary to store all synonyms corresponding to respective appearing words in the synonym dictionary and whether the word is a synonym or not can be accurately decided while considering the meansings of individual nouns constituting the word.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a technology for comparing two words and recognizing whether the two words express the same concept.

More specifically, the present invention relates to a device for determining the degree of semantic matching between words written in character codes when a document written in a natural language is processed by a computer, and relates to an automatic keyword generation device. It can be used by incorporating it into a correspondence search device, etc.

For example, in a document search, a document database creator expresses the subject of a document using words called keywords, and a searcher searches by replacing the concept of the information he/she wants to search with the keyword. Therefore, the keywords assigned by the database creator and the keywords used by searchers when searching are
If the strings match, the search is successful.

However, the words (keywords) given by the creator and the searcher
Even if they represent the same concept, they may not necessarily match as character strings due to the existence of synonyms or discrepancies in word notation. In such a case, a word matching device is required that recognizes that two words express the same concept.

[Conventional technology]

In conventional word matching devices, the following method is used to check whether words given by character codes are synonymous.

(1) A method of referring to synonyms written in a dictionary.

This is a method in which all words that are synonymous are written in a dictionary, and whether or not words are synonymous is determined by referring to the dictionary.

■ A method that determines whether or not the strings are synonyms based on the degree of matching.

This method performs matching according to the degree of matching at the character string level within a word, without particularly considering the meaning of individual nouns.

(Problems to be Solved by the Invention) However, the conventional techniques 0) and 2 above have the following problems.

As mentioned above, the method ('l) is a method of storing all synonyms in a dictionary. This method is effective when the word to be matched is a short-unit noun (the smallest meaningful unit of a Japanese word). However, this is not necessarily effective if the word is a noun continuous compound word formed by combining short unit nouns. The reason for this will be explained below. Generally, in the field of science and technology, noun continuous compounds formed by combining short nouns, prefixes, and suffixes are used as science and technology terms.

In such cases, there is a fluctuation in the number of synonyms written in each short unit. ■When short unit nouns are combined, ′
, “/′° and other symbols and prefixes.

Suffixes etc. are inserted between nouns to form derived compounds. ■Compared to the original compound word, an abbreviated word may be missing some of the character strings in the short noun, or the short noun itself may be omitted. These ■, ■, ■
Due to the causes and combinations thereof, there are a large number of noun continuous compound words (synonyms) that express one concept.It is important to cover all these synonyms and write them in a dictionary. It also requires a large amount of man-hours, and the computer's storage capacity also increases by the number of synonyms.

Furthermore, in the method (2), problems such as an increase in dictionary storage capacity and dictionary maintenance can be avoided. However, in order to match at the character string level, synonyms that are different as character strings but have the same meaning (for example, "li secondary battery", "li secondary battery", "lithium secondary battery") The disadvantage is that it is not possible to determine

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to solve the problems of the prior art described above, and to provide a word matching device that does not require a large storage capacity of a dictionary, is easy to create and maintain a dictionary, and can perform word matching with high accuracy. shall be.

(Means for solving problems) FIG. 1 is a block diagram of the principle of the present invention.

The parting section 100 selects one of the words to be matched (hereinafter referred to as
Word A) is divided into short nouns, which are the smallest meaningful words in Japanese.

The synonym replacement unit 200 replaces part or all of a word with a synonym using a synonym dictionary.

The determining unit 30G selects the word to be matched and the word that includes a synonym in part or all as the other word to be matched (hereinafter, word B).
) to determine whether the words are synonymous.

[Effect]

The single HA is divided into eight short words by the dividing section 100. For example, when word A is composed of short unit nouns at T a2 and a3, it is divided into aa2 and a3, respectively. The synonym substitution unit 200 uses an internal synonym dictionary to search for synonyms of short unit nouns. for example,
Assume that aI' is one of the six same words of at, and a3' is a synonym of a3. As a result, for word A, a,'
, al , a3 and at , al ,
A synonym candidate a3' and a synonym candidate al'+azsa3' are obtained. Here, preferably, a long unit noun is created by combining a plurality of short unit nouns, and synonyms of the long unit noun (for example, al and a+', a2' for al) are obtained from a synonym dictionary.

Words that include synonyms in part or in whole, created by the synonym replacement unit 200 as described above (the above example F is al'
al, a3': aI, al, a3':
aI, al, ax') and word A by the determining unit 300
and compare it with word B. For example, if the matching degree as a comparison result is equal to or higher than a predetermined equivalence value, the determination unit 300 determines that the word A
A comparison result indicating that and B are synonyms is output. For example, even if the comparison between the word A consisting of aI*a2*a3 and the word B does not reach a predetermined threshold, al', az.

If the word consisting of a31 exceeds a predetermined threshold,
It is determined that single HA and B are synonymous.

In this way, there is no need to store all synonyms for each word in a synonym dictionary, and it is possible to store not only the string of the entire word but also the meaning of the individual nouns that make up the word. Taking this into account, it is possible to accurately determine whether or not they are synonyms.

〔Example〕

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

FIG. 2 is a block diagram showing the hardware configuration of one embodiment of the present invention. In the figure, an input device 1 reads words written in character strings. The input device 1 may be configured with a keyboard, for example, or may be configured with an interface circuit when the illustrated hardware configuration is connected to a host computer or the like and a character string is supplied from there.

The output device 2 outputs the result of the comparison to determine whether or not the two words are synonymous, and is composed of, for example, a display or a printer. Central processing unit (hereinafter simply CP)
JJ) 3 executes a word matching program.

This program, which will be explained in detail later, includes a separation process, a short unit replacement process, a long unit @ conversion process, and a -political change calculation process.

The word table 4 stores words to be matched given from the input device 1. The short unit table 5 stores the dividing result of one word to be compared. That is, the short unit table 5 stores the results of dividing one of the words to be compared into minimum unit words that have meaning in Japanese. The candidate table 6 stores words obtained by replacing part or all of a word with a synonym. The program memory 7 stores a word matching program executed by the CPU 3, and is composed of, for example, a ROM.

The working memory 8 is a working memory used when executing a word matching program. This working memory 8 and the word table 4. The short unit table 5 and candidate table 6 are developed, for example, in a RAM memory space. The Japanese dictionary 19 is a Japanese dictionary consisting of short unit nouns, prefixes/suffixes, and symbols, and is configured on, for example, a hard disk.

The synonym dictionary 10 is a synonym dictionary that describes words having a synonym relationship, and is configured, for example, on a hard disk.

FIG. 3 is a functional block diagram of the embodiment shown in FIG. 2. The word table 4 includes a word table A11 that stores one word to be compared and a word table 812 that stores the other word.

The CPU 3 has a parting section 16. Short unit replacement section 17, good unit replacement section 18. and - It is configured with the functions of a political change calculation section 19. The parting section 16 refers to the Japanese dictionary 9, divides the words stored in the word table A11 into short noun units, and stores the nouns in the short unit table 5.

The short unit replacement unit 17 searches the short unit synonym dictionary 20 in the s1 dictionary 110 for each short unit using the short unit noun in the short unit table 5 as a key. Then, if the same II word exists in the short unit noun, the short unit replacement unit 17 generates a word in which the short unit noun in the short unit table 5 is replaced with a synonym, and replaces the original word (word words in table A11)
It is also stored in candidate table 1 indicated by reference number 13 in candidate table 6. The civil position substitution unit 18 creates a civil position noun by combining a plurality of short unit nouns in each word in the candidate table 1, and uses this as a key to create a civil position synonym dictionary 10.
Search for civil position synonym dictionary 1121 inside. If a Jf51 meaning word exists in a civil position noun, the civil position substitution unit 18 generates a word by replacing the civil position noun with the same @ word. Then, the civil position replacement unit 18 converts the candidate table 1
The words inside and the words replaced with synonyms for each unit noun are stored in the candidate table 2 indicated by reference number 14.

- The political change calculation unit 19 calculates the degree of matching between each word in the candidate table 2 and the word in the word table 812 in the word table 4, and applies the result to the word matching table 1.
5 (for example, on the working memory 8 in FIG. 2). The degree of matching is calculated using (Equation 1).

Then, if the maximum value of the degree of matching is greater than or equal to a predetermined threshold, it is determined that the words are synonymous.

(Degree of Matching) - Next, the operation of this embodiment will be explained with reference to the processing process shown in FIG. In the illustrated example, the two words to be matched are "high-speed and labor-saving search device."

The "high-speed labor-saving search system" is stored in the word table A11 and the word table 812, respectively. Among these, the word A is divided into short unit nouns,
It is divided into prefixes, suffixes, and symbols, and the divided results are stored in the short unit table 5.

Next, the short unit replacement unit 17 first searches the short unit synonym dictionary 20 containing synonyms for each short unit noun using the short unit noun in the short unit table 5 as a key. Here, as shown in FIG. 5, the short unit synonym dictionary 1120 stores headwords (short unit noun keywords) and synonyms of the headword in association with each other. From word A, it is specified that "system" is a synonym for the short unit noun "device". Then, the short unit replacement unit 17 generates a word by replacing the short unit noun "apparatus" with the synonym "system", that is, "high-speed/labor-saving search system". Then, the short unit replacement unit 17 replaces the original word and the replaced word with
Store in candidate table 1.

Next, the civil position substitution unit 18 creates a civil position noun by combining a plurality of short unit nouns, and searches the civil position synonym dictionary 1B using this civil position noun as a key.

Here, civil rank is the same! As shown in FIG. 6, the I-word dictionary 21 stores headwords (civil position noun keywords) and synonyms of the headword in association with each other. Illustrated civil rank ai
In the example of the m dictionary 21, there is no synonym for the civil position in unit 1)A. Therefore, as shown in FIG.
and the contents of candidate table 2 are the same. Then, the political change calculation unit 19 converts the words of the candidate table 2 into the word table B1.
The degree of matching is calculated according to the above equation (1) for the dish with word B of No. 2.

For example, [high-speed/labor-saving search device] in candidate table 2.
In the case of matching with word 8, this and 11 word table B12
The number of matching characters between "High-speed labor-saving search system" is "
There are 6 types: ``High'', ``Speed'', ``Relief'', ``Riki'', ``Detection'', and [Search J, both of which are composed of 10 characters, so
The degree of coincidence is 0.36. On the other hand, in the case of matching the word ``high-speed/labor-saving search system'', the degree of matching is found to be 0.83 in the same manner. For example, set the matching degree threshold to 0.
When it is set to 8, the matching value of the "high-speed, labor-saving search system" exceeds this threshold at 0.83. Therefore, it can be determined that mA and 8 are synonymous words.

FIG. 7 is a diagram showing another process.

The two words to be matched, “miniature synchrotron radiation facility’” and “miniature SOR facility,” are respectively @word table A.
11. It is stored in the word table B12. The single mA is divided into short unit nouns, prefixes/suffixes, and symbols by the dividing unit 16, and the results are stored in the short unit table 5. Next, the short unit replacement unit 17 first searches the short unit synonym dictionary of FIG. 5 using the short unit noun of the short unit table 5 as a key. Here, since there is no synonym for the short unit to be replaced, the contents of short unit table 5 and candidate table 1 are the same. Next, the long unit substitution unit 18 creates a civil position noun by combining a plurality of short unit nouns, and uses this civil position noun as a key to search the civil position synonym dictionary shown in FIG. 6.

Then, a word is generated by replacing the Nagasato noun with a synonym. Then, the original word and the replaced word are stored in the candidate table 2. - The political change calculation unit 19 calculates the degree of matching between the words in the candidate table 2 and the word table 812, and stores the result in the word matching degree table 15. Then, a threshold value (for example, 0.8) is added to the word matching table 15.
A combination of the above (“compact SOR facility” and “compact SOR facility”)
Since "OR facility") exists, it is determined that they are synonyms.

(Effects of the Invention) As explained above, in the present invention, words to be matched are divided into short nouns, part or all of the words are replaced with synonyms using a synonym dictionary, and then By performing column matching, it is determined whether the original words are synonymous. Therefore, it is not necessary to write all of the synonyms for each word that appears in the dictionary, so it is possible to avoid the problems of increasing the storage capacity of synonym dictionaries and increasing the man-hours of dictionary maintenance. This is effective in that it is possible to perform word matching that takes into account not only the columns but also their meanings.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of the principle of the present invention, Fig. 2 is a block diagram of the hardware configuration of an embodiment of the present invention, Fig. 3 is a functional block diagram of the embodiment shown in Fig. 2, and Fig. 4 is a block diagram of the embodiment of the present invention. FIG. 5 is a diagram showing an example of the short unit synonym dictionary 120, FIG. 6 is a diagram showing an example of the civil unit synonym dictionary 21, and FIG. 7 is a diagram showing the processing process of the embodiment of the invention. It is a figure which shows another processing process of an Example. 1... Input device, 2... Output device, 3... CPU
, 4... Word table, 5... Short unit table, 6
...Candidate table, 7.Program memory, 8.
...Working memory, 9...Japanese dictionary, 10...Synonym dictionary, 11...Word table A112...Word table B113...Candidate table 1.14...Candidate table 2.15 ...Word matching table, 16...
- Breaking section, 17...Short unit replacement section, 18...
Civil rank replacement department, 19... - Political change calculation department, 20... Short unit synonym dictionary, 21... Civil rank synonym dictionary, 100
... Separation section, 200 ... Synonym replacement section, 30
0... Judgment section. Figure 1 is a block diagram of the principle of the present invention. Figure 2 is a block diagram of the hardware configuration of an embodiment of the present invention.
F! J Functional block whistle 1M of the embodiment shown in FIG.

Claims (1)

[Claims] A separating section that divides one word to be matched into short nouns, which are the smallest meaningful words as Japanese words; A synonym substitution unit that replaces words with synonyms, and a word to be matched and a word that partially or completely includes a synonym are compared with the other word to be matched to determine whether or not the words are synonyms. A word matching device comprising: a determining unit that makes a determination.