JP3360261B2 - Natural language analyzer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a natural language analyzer capable of increasing the success rate of information retrieval and shortening the time required for information retrieval in information retrieval using a company name or the like.

[0002]

2. Description of the Related Art Hitherto, in information retrieval, a search was performed by inputting a part of a term for a name such as a company name or the like according to a specific rule or custom. Hereinafter, a description will be given of information retrieval by company name as an example. Common uses of an information retrieval system using a company name character string as a search key include a corporate name consisting of compound nouns (words such as place names, occupation names, and product names appearing at the beginning of the company name) Often, endings (words appearing at the end of a company name, such as a center, a school, etc.) or words located in the middle are abbreviated under specific rules and customs. Hereinafter, a name composed of a degenerated character string that substitutes for such a character string will be referred to as a “replacement name”. By using both the company name and the replacement name as a search key for the company name having the replacement name, a flexible search can be performed. Hereinafter, the process of outputting a replacement name for a search key composed of a character string is referred to as “derivation of a search key”. Conventionally, as a means for deriving a search key of a company name, there has been a process of deriving a search key using a table. In the process of deriving a search key using a table, place names and occupation names that are often used as titles are registered in the table, and when configuring the database, they are registered in the table in order from the top of the company name to be searched. A method of deriving a character string obtained by deleting a part that matches the character string used as a search key was used.

[0003]

As described above, the conventional method of deriving a retrieval key by referring to a table has the following problems. These specific examples will be described with reference to FIG. In FIG. 5, the upper part shows a correct method of extracting a substitute name from a search key, for example, in Example 5-1, a state where "doll" is derived as a substitute name from "cafe doll". At the bottom, a conventional method of deriving a search key using a table is shown. (Problem 1) Even if a character string is not a word constituting a company name, if it is registered in a table, it is deleted, so that an inappropriate search key may be derived. That is, as shown in an example 5-1 in FIG. 5, it is assumed that there are two heading names “cafe” and “caffedo” corresponding to the company name “café doll” in the heading name table in a forward match. .
In the method of deriving the search key by referring to the table, no matter what the character string after deleting the caption name, the character string with the larger number of characters is deleted as the caption name. A missing string will be derived as a search key.

(Problem 2) When the title is composed of a compound noun, an inappropriate search key may be derived because only the word at the head position registered in the table is deleted. That is, as shown in Example 5-2 of FIG. 5, even if both "Interior" and "Gallery" are registered in the title table with respect to the company name "Interior Gallery Takahashi", retrieval by referring to the table is performed. In the key deriving means, only the "interior" at the beginning is deleted, and "gallery Takahashi" is derived as a search key. (Problem 3) A word located in the middle of a compound noun cannot be deleted. That is, as shown in examples 5-3 and 5-4 in FIG. 5, the search keys derived from the company names "Hachioji Agricultural Cooperative" and "Katsumi Takahashi Lawyer Office" are "Hachioji Agricultural Cooperative" and "Takahashi Lawyer Office" respectively. Is appropriate, but in the means of deriving a search key by referring to a table, only the character string at the beginning is subject to deletion, so in the case of "Hachioji City Agricultural Cooperative", an inappropriate search key "City Agricultural Cooperative" is derived However, in the case of "Katsumi Takahashi Lawyer Office", a search key cannot be derived. An object of the present invention is to solve these conventional problems and to derive a search key that is easy to be input as a search condition for general use in an information search system using a compound noun as a search key, As a result, it is an object of the present invention to provide a natural language analyzer capable of increasing the success rate of information retrieval and shortening the time required for information retrieval.

[0005]

In order to achieve the above object, a natural language analysis apparatus according to the present invention comprises an analysis dictionary (12) in which words and meanings of the words are registered. In a device for analyzing a character string consisting of words of
Analysis means (14) for dividing the input character string into words by referring to the analysis dictionary (12) and assigning a semantic classification to the divided words, and a meaning corresponding to each of the divided words A semantic pattern table (13) for registering various combinations of classifications and defining rules for degenerating character strings in word units, and the analyzing means (14) refers to the semantic pattern table (13). Thus, a degenerated character string that substitutes for a character string is generated and output.

[0006]

According to the present invention, an analysis dictionary for registering words constituting a company name and their meanings, and a meaning pattern table for registering a derived portion corresponding to the meaning pattern of the words constituting the company name are prepared in advance. A search derived by dividing a character string consisting of compound nouns such as company names to be searched into words using an analysis dictionary, adding a semantic classification to each word, and referring to a semantic pattern. Automatically determine key parts. As a result, (a) it is possible to derive a search key by deleting a character string from a single word constituting a compound noun. (B) By referring to the meaning pattern added to a word, it is possible to determine a combination of words that is appropriate as a derived search key.

[0007]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a configuration diagram of a search key derivation system showing one embodiment of the present invention. The company name search key derivation system includes an input unit 11 for inputting a company name character string serving as a search key, an analysis dictionary 12 in which words forming the company name and their meanings are registered, and the meaning of the words forming the company name. A derived pattern corresponding to the pattern is registered, and a semantic pattern table 1 in which rules for reducing a character string in word units are defined.
3, an analysis unit 14 for deriving a search key for an input character string with reference to the dictionary 12 and the table 13, and an output unit 15 for outputting the derived character string. In the present invention, the analysis unit 14 uses the analysis dictionary 12 to divide a character string input from the input unit 11 into words. Then, a semantic classification is given to each of the divided words. Next, various combinations of the meaning classification corresponding to each word are registered in the meaning pattern table 13. Using these analysis dictionaries 12 and the semantic pattern table 13, a vertical character string that substitutes for a character string is derived.

FIG. 2 is a diagram showing the contents of the analysis dictionary and the meaning pattern table in FIG. In the analysis dictionary 12, words constituting the company name and the meaning corresponding to each word are registered. The meaning is a classification name based on the attributes of the words that make up the company name, such as "proper name" such as Fuji and Sanwa, "ending" such as the center, or "professional name", "verb", " Address, Local name, Subordinate organization
Etc. The meaning pattern in the meaning pattern table 13 is a type of a series of meanings of the words constituting the company name, and the derived part is to specify a derived search key by a method of combining the divided words. . In the semantic pattern table, a semantic pattern having a strong statistical correlation with the derived part is registered together with the corresponding derived part.

FIG. 3 is an operation flowchart of the retrieval key deriving system showing one embodiment of the present invention. First, in a word division step 31, the company name character string input from the input unit 11 is divided into words by referring to the analysis dictionary 12. Next, in a meaning pattern generation process P32, a meaning pattern registered in the analysis dictionary 12 is added to the words forming the company name to generate a meaning pattern of the words forming the company name. Next, a derived part determination process P33
Then, referring to the semantic pattern table 13, if the semantic pattern of the word constituting the company name is registered in the table 13, it is determined that a derived search key exists, and the corresponding derived part is obtained. . Lastly, if there is a derived part, in the search key generation step 34, the words constituting the company name are combined in accordance with the combination of the words specified in the derived part, thereby being derived from the input company name. Generate a search key.

FIG. 4 is a diagram showing a specific example of the retrieval key deriving process according to the present invention. The upper part of FIG. 4 shows a case where a replacement name is extracted from a search key. The lower part of FIG. 4 shows a method of deriving a search key according to the present invention in the order of word division, meaning pattern, derived part, and derived search key. This shows a state in which the search key is correctly derived. Example 4 in FIG.
FIG. 1 shows an example in which a plurality of words registered in a dictionary exist at the beginning. In the company name "Cafe Doll"
"Cafe (professional name)", "Cafedo (professional name)", and "Doll (general noun)" are words registered in the dictionary, and "le" is not registered, so the words "cafe" + "doll" Divided. This meaning pattern is "occupation name + common name"
And the corresponding derivative part is "second word",
"Doll" is derived as a search key. As described above, in the present invention, since the word is divided using the name analysis dictionary 12, it is possible to derive a search key by deleting a character string in units of words constituting a compound noun.

In the example 4-2 of FIG. 4, the title is composed of a compound noun. In the example 4-3 and the example 4-4, the words located in the middle of the compound noun are deleted to be replaced. An example of deriving the name is shown. That is, in Example 4-2, the meaning pattern of the company name "Interior Gallery Takahashi" is "product name + ending + unique name (surname)", and the derived part corresponding to this meaning pattern is "third word". Therefore, "Takahashi" is derived as a search key. Example 4
In -3, the meaning pattern of the company name "Hachioji Agricultural Cooperative"
Since "place name + place name suffix + occupation name" and the derived part corresponding to this semantic pattern is "first word + third word ", "Hachioji agricultural cooperative" is derived as a search key. Example 4
In -4, the meaning pattern of the company name "Katsumi Takahashi Lawyer Office" is "proper name (surname) + proper name (first name) + occupation + ending"
And the derived part corresponding to this semantic pattern is “1st
Since “word + third word ”, “Takahashi Lawyer Office” is derived as a search key. In this way, by referring to the meaning pattern added to the word, it is possible to determine a combination of words that is appropriate as a derived search key.

[0012]

As described above, according to the present invention,
In an information retrieval system using a compound noun as a retrieval key,
In general, it is possible to derive a search key that is easy to input as a search condition when used. In particular, compound nouns
By omitting the middle part, the success rate of information search can be increased, and the time required for information search can be reduced.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a company name search key derivation system showing one embodiment of the present invention.

FIG. 2 is a diagram showing a specific example of an analysis dictionary and a meaning pattern table in FIG. 1;

FIG. 3 is a processing flowchart of a company name search key derivation system according to the present invention.

FIG. 4 is a diagram showing a specific embodiment of a company name retrieval key derivation system according to the present invention.

FIG. 5 is a diagram showing a specific example of a search key deriving process using a conventional table.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 11 Input part 12 Analysis dictionary 13 Meaning pattern table 14 Analysis part 15 Output part 31 Word division process 32 Meaning pattern judgment process 33 Derived part judgment process 24 Derived search key generation process

Continuation of the front page (56) References Iwase, A., Takahashi, K., Advanced search method of company name, IPSJ 45th Annual Meeting, IPSJ, October 11, 1992 Vol. 3, pp. 3-245 to 3-246 (58) Fields investigated (Int. Cl. ⁷ , DB name) G06F 17/30

Claims (1)

(57) [Claims] 1. An apparatus for analyzing a character string composed of one or more input words, comprising an analysis dictionary in which words and meanings of the words are registered. Analysis means for dividing the words by reference and assigning a semantic classification to the divided words, and a sentence derived from a semantic pattern corresponding to each of the divided words
Register the pair of combinations of string portion; and a meaning pattern table for degenerating string word by word, the above analyzing means beginning of the character string by referring to the meaning pattern table, the intermediate , Or a combination of these
A natural language analysis apparatus for generating and outputting a degenerated character string omitting a false character string.