JPH07334503A - Natural language processing system - Google Patents

Abstract

PURPOSE:To transfer a compound noun having the modification of a word, which is required for sentence structure analysis, and meaning information on words, which are required for meaning analysis, and also, to provide a comparatively simple data structure. CONSTITUTION:In the natural language processing system which morphemeanalyzes an input sentence and analyzes the sentence structure and the meaning from the analyzed result, data obtained in morpheme analysis is set to be word data obtained by dividing the input sentence into respective word units and describing attribute data following the respective words, and compound noun data obtained by describing meaning information on respective nouns constituting the compound noun among the words. In the meaning analysis of the compound noun, meaning information on the respective nouns constituting the compound noun are obtained by referring to compound noun data. Common compound noun data is set to be a retrieval/reference object for the compound noun of the same notation.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a natural language processing system, and more particularly to data processing for passing compound noun data obtained by morphological analysis processing to syntax / semantic analysis processing.

[0002]

2. Description of the Related Art Natural language processing using a computer such as a word processor, machine translation, document database, and hypertext has been put into practical use.

In natural language analysis for this purpose, a sentence to be analyzed is first divided into morpheme units (minimum units of word structure), and morpheme analysis is performed to clarify the properties of each morpheme. After this, a syntactic analysis that analyzes from the syntactic rules of natural language, and then a semantic analysis that removes ambiguity and vagueness,
Perform contextual analysis.

In the syntactic analysis, it is determined whether or not the morphologically analyzed sentence is a correct sentence by using a grammar, and when the sentence is correct, a tree structure (a parse tree) is obtained as the syntactic analysis result.

In this syntactic analysis process, since attention is paid only to the grammatical conformity, syntactic ambiguity occurs and many syntactic trees are generated. A semantic analysis process is performed to select the correct parse tree from among these.

In the semantic analysis process, not only the grammatical category of a word (corresponding to a part of speech) but also its semantic information is used.

Here, the morphological analysis is processed by a program described in C language, and the syntax / semantic analysis is performed by Pr.
Processing is performed by a program written in the log language.

Therefore, an interface means is required for passing the data obtained as the morphological analysis result to the syntax / semantic analysis processing. The passing data at this time is
The input character string is divided into words, and the data in the format that describes the attribute data attached to each word (hereinafter, "d"
ict data ”).

[0009]

When there is a compound noun in the input character string, it is divided into individual nouns by morphological analysis, but the dict data treats the compound noun as one word data.

This is because it is rational to check the dependency between words before and after it, using a compound noun as one noun because the dependency of words is checked at the stage of parsing.

However, at the stage of semantic analysis, it is necessary to know what kind of noun the compound noun is composed of and what kind of dictionary information each noun had (including semantic ones). .

At present, the dict data of compound noun is
Since only a limited amount of information that the original noun had was inherited, the accuracy of semantic analysis of compound nouns was deteriorated.

It is an object of the present invention to provide a natural language processing system capable of accepting a word necessary for syntactic analysis and passing a compound noun carrying the semantic information of a word necessary for semantic analysis.

Another object of the present invention is to provide a natural language processing system capable of passing data with a relatively simple data structure.

[0015]

According to the present invention, in order to solve the above problems, a morpheme analysis is performed on an input sentence, and a natural language processing system for performing syntactic analysis and semantic analysis from the analysis result is obtained by the morpheme analysis. The data is divided into input words into word units, and word data describing attribute data associated with each word, and compound noun data describing semantic information of each noun forming a compound noun of the words, The syntactic analysis generates a syntactic tree from the word data, and the semantic analysis retrieves the compound noun data by using the notation as a key for compound nouns having the same notation among the word data and semantic information of each noun. It is characterized in that

[0016]

[Function] In addition to the conventional word data as the morphological analysis result data, compound noun data is generated, and the semantic information of each noun forming the compound noun is referred by referring to the compound noun data for the semantic analysis of the compound noun. obtain.

For compound nouns having the same notation, common compound noun data is searched and referred to.

[0018]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows an embodiment of the present invention and shows the structure of data passed to a syntax / semantic analysis process by a morpheme analysis process.

As morphological analysis results for the input sentence, semdict data as compound noun data is generated in addition to dict data as word data, and data is passed for syntax / semantic analysis.

As in the conventional case, the dict data has a data structure including the part of speech and notation of a word, grammatical information and semantic information.

Semdict data (compound noun data)
Is a data structure including a compound noun notation of a word to be a compound noun, an original noun notation, and semantic information of the original noun of each word of the dict data.

For example, an input sentence "experimental result is subjected to machine processing" is divided into morphemes, and the result is "experiment / result / a / machine / processing / to / multiply / multiply". The dict data that becomes the word data of this morpheme is dict (noun, experimental result, ~, Sem1) dict (case particle, 〜, ~) dict (noun, machine processing, ~, Sem2) dict (case particle, , ~, ~) Dict (verb, call, ~, ~). Here, Sem is semantic information that each noun has, for example, a semantic feature or a thesaurus code.

The semdict data at this time is semdict (compound noun, experimental result, [[experiment, Se
m11], [result, Sem12]]) semdict (compound noun, machine processing, [[machine, Se
m21], [processing, Sem22]]).

Such dict data and semdic
In the syntax / semantic analysis process delivered with the t data, the syntax analysis generates a syntax tree from the dict data by using word modification.

In the semantic analysis for determining the semantic compatibility of the generated syntax tree, the word "experimental result" in the dict data is not in the dictionary, so that the meaning information "Sem1" is nothing. Not described.

Therefore, it is searched whether or not the compound noun expression is included in the semdict data, and the semantic information Sem of "experiment" is searched.
11 and the semantic information Sem12 of "result" are referred to. This reference can be used for semantic analysis of the compound noun “experimental result”.

Similarly, with respect to the compound noun "machine processing", the notation can be used as a key to search the semidict data and obtain the semantic information.

Therefore, according to this embodiment, dict
Among the data, the semantic information of the compound noun can be referred to and used at the stage of the semantic analysis by referring to the semidict data to see what kind of noun the compound noun is composed of and what kind of semantic information each noun has.

Moreover, since the conventional word data (dict data) is not changed at all, the syntactic analysis process need not be changed.

The semdict data can be used as common semantic information for compound nouns having the same notation. That is, when the semantic information for each word of the compound noun is described in the dict data, when a plurality of compound nouns having the same notation appear in one input sentence or a plurality of input sentences, the same meaning information is included in each dict data. Would result in redundant data. In this respect, in the present embodiment, by passing one semdict data to the compound noun having the same notation, the meaning information can be obtained by referring to the same semdict data when the compound noun having the same notation appears. Can be obtained.

[0031]

As described above, according to the present invention, as the data of the morphological analysis result, in addition to the conventional word data, compound noun data describing the semantic information of each noun forming a compound noun of words is used. Since the compound noun is generated and passed to the syntactic / semantic analysis and the compound noun is searched for compound noun data using the notation as a key to obtain the semantic information of each noun, the following effects are obtained.

(1) The accuracy of the semantic analysis of the input sentence including the compound noun can be improved.

(2) No change is required in the syntactic analysis process even when the compound noun data is added.

(3) With regard to the transfer data, it is sufficient to add the compound noun data to the conventional word data, and the data amount hardly increases.

(4) For compound nouns having the same notation, common compound noun data can be commonly used as a search / reference object.

[Brief description of drawings]

FIG. 1 is a transfer data structure showing an embodiment of the present invention.

Claims (1)

[Claims] 1. A natural language processing system for performing morphological analysis of an input sentence and performing syntactic analysis and semantic analysis from the analysis result, wherein the data obtained by the morphological analysis divides the input sentence into word units and Word data describing the accompanying attribute data, and compound noun data describing the semantic information of each noun forming a compound noun of the words, the syntactic analysis generating a syntactic tree from the word data, The natural language processing system wherein the semantic analysis obtains semantic information of each noun by searching the compound noun data for the compound nouns having the same notation among the word data by using the notation as a key.