JP2003281137A - System and method of processing natural language, and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To accurately determine whether an inputted sentence is an interrogative sentence or not without depending solely on extraction of an interrogative. <P>SOLUTION: A word belonging to a case category W corresponding to the interrogative in an inputted sentence is searched, and whether there is a case category P corresponding to a postpositional particle which has a function of offsetting the effect of the word (w) of making the whole sentence interrogative just after the found word (w) is determined. Then, the largest possible phrase to make the range of the function of the word (w) for requiring the inputted sentence to be interrogative largest is extracted. When the largest phrase extends over the whole sentence, the interrogative sentence is of the five-Ws-and-one-H type; and when the largest phrase doesn't extend over the whole sentence, the interrogative sentence is of the Yes/No type. <P>COPYRIGHT: (C)2004,JPO

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 a natural language processing method for mathematically handling a natural language used by humans for daily communication.
The present invention also relates to a computer program, and more particularly to a natural language processing system and a natural language processing method for performing syntactic and semantic analysis of Japanese syntax, and a computer program.

More particularly, the present invention relates to a natural language processing system and a natural language processing method for performing syntactic / semantic analysis of Japanese syntax according to grammatical rules, and a computer program. In particular, an input sentence is a question sentence. The present invention relates to a natural language processing system, a natural language processing method, and a computer program.

[0003]

2. Description of the Related Art Words used by humans for daily communication such as Japanese and English are called "natural language". Many natural languages have a naturally occurring origin and have evolved with the history of humanity, ethnicity, and society. Of course, a person can communicate by gesturing or gesturing, but natural language can realize the most natural and advanced communication.

On the other hand, with the development of information technology, computers have become established in human society and have been deeply permeated in various industries and daily lives. Now, not only computer data but almost all information contents such as images and sounds are handled on the computer, and it is possible to perform advanced processing such as editing / processing, storage, management, transmission and sharing of information. .

Natural language has an abstract and ambiguous nature by nature, but by treating sentences mathematically,
Computer processing can be performed. As a result, various applications related to natural language such as machine translation, dialogue system, search system, etc.
The service is realized.

Natural language processing is generally divided into processing phases of morphological analysis, syntactic analysis, semantic analysis, and context analysis.

In the morphological analysis, a sentence is segmented into morphemes, which are the smallest semantic units, and a part-of-speech recognition process is performed. In the syntactic analysis, a sentence structure such as a phrase structure is analyzed based on grammatical rules. Since the grammar rules have a tree structure,
The result of parsing is generally a tree structure in which individual morphemes are joined. In the semantic analysis, based on the meaning (concept) of the words in the sentence and the semantic relationship between the words, the semantic structure expressing the meaning conveyed by the sentence is obtained, and the semantic structures are synthesized. In context analysis, a sentence (discourse), which is a series of sentences, is regarded as a basic unit of analysis, and a discourse structure is constructed by obtaining a semantic unity between sentences.

By the way, when constructing a dialogue system or a question answering system using natural language processing, whether the input from the user is a regular question sentence according to the grammar or a non-sentence sentence, a Yes-No type is used. The process of determining whether an interrogative sentence or a 5W1H type interrogative sentence is extremely important in generating a response to a user input.

In the current syntax / semantic analysis system, only a technique of identifying a question word in a question sentence and generating an answer corresponding thereto is used. in this case,
For example, in response to user input, "Does anyone know the word digital divide?"
It tries to find out who knows the word digital divide from the knowledge base by identifying the interrogative word "who".

However, the user input in this case does not actually ask who knows the word digital divide, but rather, for example, what percentage of Japanese people know the word. Probably listening.

This is a phenomenon that occurs when an interrogative in a question sentence is processed by limiting the meaning usage to one.
As a result, the intention of the sentence is not correctly understood.

[0012]

An object of the present invention is to provide an excellent natural language processing system, natural language processing method, and computer program capable of suitably performing semantic analysis of Japanese syntax according to grammatical rules. To do.

A further object of the present invention is to provide an excellent natural language processing system, natural language processing method, and computer program capable of accurately determining whether or not an input sentence is a question sentence. Especially.

A further object of the present invention is to provide an excellent natural language processing system and natural language capable of accurately determining whether or not an input sentence is a question sentence, without simply relying on extraction of question words. It is to provide a processing method and a computer program.

A further object of the present invention is whether the input sentence is a regular interrogative sentence or non-sentence sentence according to the grammar, Ye.
An object of the present invention is to provide an excellent natural language processing system and natural language processing method, and a computer program capable of determining whether the question sentence is an s-No type or a 5W1H type question sentence.

[0016]

The present invention has been made in consideration of the above problems, and the first aspect thereof is a natural language for determining whether an input sentence is a question sentence or not. A processing system or a natural language processing method,
A first means or step of determining whether or not the input sentence includes a word belonging to a part-of-speech category W corresponding to an interrogative, and the word w is the entire sentence immediately after the word w found by the first means. Natural language processing system or natural language processing comprising: a second means or step of determining whether or not there is a part-of-speech category P corresponding to a particle having a function of canceling the effect of making the question sentence Is the way.

A second aspect of the present invention is a natural language processing system or natural language processing method for determining whether an input sentence is an interrogative sentence, wherein the part of speech corresponding to the interrogative phrase is included in the input sentence. A first means or step for determining whether or not a word belonging to category W is included;
Second means or step for determining whether or not there is a part-of-speech category P corresponding to a particle having a function of canceling the effect of making the whole sentence into a question sentence immediately after the word w discovered by the means of Third means or step for extracting the maximum possible phrase which is the maximum range covered by the function of requesting that the word w is an interrogative sentence in the input sentence, and the maximum phrase extracted by the third means is the entire sentence. Fourth means or step for determining whether or not the input sentence is a question sentence depending on whether
A natural language processing system or a natural language processing method characterized by comprising:

In the natural language processing according to the present invention, it is preferable to treat a sentence described in a structure in which a relative phrase is understood as an input. More specifically, such an input sentence is an LFG.
(Lexical-Functional Grammar) f-str which is the result of parsing and semantic analysis based on grammar theory
It is a sentence described in the structure format.

Further, the third means or step is a maximum phrase that can be a verb phrase related to the word w and a verb phrase related to the noun phrase to which the relative verb is related when the word w is included in the relative verb clause. To extract. The input sentence is f-structure
If the sentence is written in the form, the part of speech for each morpheme has been certified, and the dependency relation is clear, so f
By describing a grammar rule that a predetermined INT mark is embedded in -structure, the maximum phrase can be easily extracted. That is, first, the INT mark is added to the question word, and the INT
The mark propagates to the upper part of the sentence to the maximum extent.

When it is determined by the fourth means or step that the maximum phrase extends to the entire sentence, it is further determined whether the entire input sentence is a question sentence. And
If the determination result is affirmative, it is determined that the input sentence is a 5W1H type question sentence, and if the determination result is negative, it is determined that the input sentence is a non-grammatical sentence.

If no word belonging to the part-of-speech category W is found in the input sentence by the first means or step, the second means belongs to the part-of-speech category P immediately after the word w in the input sentence by the step. When the word is found, and when the step determines that the maximum phrase does not extend to the entire sentence, the fourth means further determines whether or not the input sentence is a question sentence. If the determination result is affirmative, it is determined that the input sentence is a Yes-No type interrogative sentence, and if the determination result is negative, it is determined that the input sentence is an affirmative sentence.

Whether or not the whole sentence is a question sentence can be determined depending on whether the end is a final particle indicating a question or a symbol indicating a question.

Further, a third aspect of the present invention is a computer written in a computer-readable format so as to execute natural language processing for determining whether an input sentence is a question sentence on a computer system. In the program, a first step of determining whether or not the input sentence includes a word belonging to a part-of-speech category W corresponding to an interrogative word, and a word w immediately after the word w found by the first step. A second step of determining whether or not there is a part-of-speech category P corresponding to a particle having a function of canceling the effect of making the entire sentence into a question sentence.

Further, a fourth aspect of the present invention is a computer written in a computer readable format so as to execute natural language processing for determining whether an input sentence is a question sentence on a computer system. In the program, a first step of determining whether or not the input sentence includes a word belonging to a part-of-speech category W corresponding to an interrogative word, and a word w immediately after the word w found by the first step. The second step of determining whether there is a part-of-speech category P corresponding to a particle having a function of canceling the effect of making the entire sentence into a question sentence, and a function of requesting that the word w is a question sentence in the input sentence Whether the input sentence is an interrogative sentence or not depending on the third step of extracting the maximum possible phrase that is the maximum range of the input sentence and whether or not the maximum phrase extracted by the third step covers the entire sentence. Is a computer program characterized by comprising a fourth step, the determining.

The computer program according to each of the third and fourth aspects of the present invention defines a computer program written in a computer-readable format so as to realize a predetermined process on a computer system. . In other words, by installing the computer program according to each of the third and fourth aspects of the present invention in the computer system, a cooperative action is exerted on the computer system, and the first and the second aspects of the present invention are realized. Two
It is possible to obtain the same operational effects as the natural language processing system and the natural language processing method according to each aspect of.

According to the natural language processing system or the natural language method of the present invention, it is possible to highly accurately determine whether or not the input sentence is a question sentence. Therefore, the question answer can be made by using the determination result of the question sentence in the dialogue system. That is, in the dialogue system, the syntax / semantics of the input sentence may be analyzed, and in response to the determination that the sentence is an interrogative sentence, an answer to the interrogative sentence may be generated as a companion sentence to the input sentence. For example, if it is determined to be a Yes-No type interrogative sentence, an answer sentence based on the Yes-No format may be generated, and if it is determined to be a 5W1H type interrogative sentence, An answer sentence corresponding to the inquired matter may be generated.

Further objects, features and advantages of the present invention are as follows.
It will be apparent from the embodiments of the present invention described later and the more detailed description based on the accompanying drawings.

[0028]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below in detail with reference to the drawings.

The natural language processing system according to the present invention judges whether or not the input sentence is an interrogative sentence. For example, in the syntactic / semantic analysis process based on the LFG (Lexical-Functional Grammar) grammar theory. It can be embedded and implemented. In LFG, the linguistic knowledge, or grammar, of native speakers is organized as a component separate from computer processing and other non-grammatical processing parameters that affect the processing behavior of the computer. First, an overview of the natural language processing system will be briefly described. Although the present embodiment will be described based on the LFG grammar theory, of course, the present invention can be similarly applied to an analysis system having another grammar rule.

FIG. 4 schematically shows the configuration of the LFG-based natural language processing system 1.

The morpheme analysis unit 2 has a morpheme rule 2A and a morpheme dictionary 2B relating to a specific language such as Japanese, and divides an input sentence into morphemes, which are the smallest semantic units, to perform a part of speech recognition process. For example, if the sentence "My daughter speaks English." Is entered, the result of morphological analysis is "I {Noun} {up} daughter {Noun} is {up} English {Noun}.
{Up} speak {Verb1} {tr} and {jp}. {pt} "is output.

The result of such morphological analysis is then input to the syntactic / semantic analysis unit 3. The syntactic and semantic analysis department
It has dictionaries such as grammar rules 3A and valence valence dictionary 3B, and analyzes the phrase structure based on grammar rules and expresses the meaning conveyed by the sentence based on the meaning of words in the sentence and the meaning relation between words. Analyze the semantic structure (The valence dictionary describes the relationship between verbs and other constituent elements in the sentence, such as the subject, and can extract the semantic relationship between predicates and related words). Then, as a result of the syntactic analysis, "c-structure (constituent structur), which represents a phrase structure of a sentence including words and morphemes as a tree structure.
e) ", and as a result of semantically and functionally analyzing the input sentence based on the case structure of the subject, object, etc., such as question sentence, past tense, polite sentence," f-structure (functional stru
cture) "is output.

5 and 6, the c-structure and f-st obtained as a result of processing the input sentence "My daughter speaks English" by the syntactic / semantic analysis unit 1.
In each of the figures, the figures are shown.

The c-structure represents the structure of words and phrases in a sentence in a tree structure format, and is defined by the syntax category. For example, a phonological interpretation for generating a phoneme string can be performed based on c-structure. On the other hand, f-structure is
It is a clear expression of grammatical functions and is composed of grammatical function names, semantic forms, and feature symbols. By referring to f-structure, the subject, the object, and the complement
t) and modifiers (adjunct) can be understood. f-structure is c-structure
It is a set of features attached to each node of ure, and is expressed in the form of an attribute-attribute value matrix as shown in FIG. In other words, the left side in [] is the feature (attribute)
And the right side is the feature value (attribute value).

For details of LFG, see, for example, R.
Paper "Lexical-Functio" co-authored by M. Kaplan and J. Bresnan
nal Grammar: A Formal System for Grammatical Repre
sentation "(The MIT Press, Cambridge (1982). Repr
inted in Formal Issues in Lexical-Functional Gramma
r, pp. 29-130.CSLI publications, Stanford Univers
ity (1995).).

Next, the question sentence determination by the natural language processing according to the present invention will be described in detail.

In the present invention, first, a part-of-speech category W corresponding to an interrogative and a part-of-speech category P corresponding to a particle having a function of canceling the effect that the interrogative makes the whole sentence an interrogative sentence are defined.

Here, the question words classified into the part-of-speech category W are "who" and "what". Particles classified into the part-of-speech category P are “ka” and “mo”. For example, as can be seen from the fact that “ka” is a subpositional particle and “mo” is a particle, the classification of the part-of-speech category P does not match the general part-of-speech classification.

FIG. 1 shows, in the form of a flowchart, a processing procedure for determining whether or not the input sentence is an interrogative sentence.

According to this processing procedure, using the part-of-speech categories W and P defined in advance, whether the input from the user is a regular question sentence or a non-sentence sentence according to the grammar, and further, Yes It is possible to determine whether it is a No type question sentence or a 5W1H type question sentence.

However, the statement input to this processing flow is
For example, it is preferable that the sentence is described in a structure in which the relative phrase is understood after the parsing process is completed.

First, it is determined whether the input sentence from the user is a sentence including a word belonging to the part-of-speech category W (step S1).

When the input sentence includes a word belonging to the part-of-speech category W (hereinafter referred to as word w), the word w is further added.
Immediately after, it is determined whether or not there is a word belonging to the part-of-speech category P (step S2).

In step S2, the determination result is negative,
That is, when there is no word belonging to the part-of-speech category P immediately after the word w, the maximum phrase allowed in the input sentence is INT.
A process of adding a mark (Interrogative) is performed (step S3).

Here, the "maximum possible phrase" is a phrase in the maximum range covered by the function of requesting that the word w found in step S1 is a question sentence. Specifically, it refers to the verb phrase to which the word w relates, and the verb phrase to which the noun phrase to which the relative word belongs, when the word w is included in the relative phrase.

For example, in the case of the input sentence "Who owns the cat?", The interrogative word "who" relates to "keep", so "INTO", that is, the entire sentence is marked with the INT mark. Become. In this case, "who owns the cat." Is not recognized as a legitimate Japanese sentence. Also,
In the case of "Who has kept the cat away?", First, "keep" is marked with INT, and the noun phrase "cat" is associated with the relative clause "who owns". The phrase "escape" (that is, the entire sentence) is also given INT. When the relative clause has a multiple nested structure, INT is recursively propagated in the upper direction of the entire sentence.

In step S4, it is determined whether INT is added to the entire sentence. When it is determined that INT is added to the entire sentence, further step S5
In, it is determined whether the entire sentence is a question sentence.
This can be determined by whether or not the end of the input sentence ends with a final particle "ka" indicating a question or a symbol indicating a question.

If the result in step S5 is affirmative, it is determined to be a 5W1H type interrogative sentence. Also,
If the judgment result is negative, it is judged that the input sentence is a non-grammatical sentence. When the input by the user is determined to be a non-grammatical sentence, the correct meaning may be asked back to the user.

On the other hand, if a word belonging to the part-of-speech category W is not found in the input sentence in step S1, a word belonging to the part-of-speech category P is found immediately after the word w in the input sentence in step S2, or step When it is determined in S4 that INT is not added to the entire sentence, it is further determined in step S6 whether the entire sentence is a question sentence. This can be determined based on whether or not the end of the input sentence ends with a final particle "ka" indicating a question or a symbol indicating a question, as in step S5.

If the result in step S5 is affirmative, it is determined to be a Yes-No type interrogative sentence. If the determination result is negative, it is determined that the input sentence is an affirmative sentence.

The processing procedure as shown in FIG.
This can be easily realized by describing a grammar rule that embeds the INT mark in f-structure, but it can also be implemented as a processing function included in the syntactic / semantic analysis unit 3 itself.

Here, the question sentence determination process shown in FIG. 1 will be specifically described by taking the following six sentences as an example.

[0053] (1) Who wrote "The Origin of Species". (2) Who wrote "The Origin of Species"? (3) Who read the book you wrote. (4) Who read the book you wrote? (5) Everyone is reading the Bible. (6) Does anyone read the Bible?

The sentence example (1) is "Y" in step S1.
es "," No "in step S2," Yes "in step S4," No "in step S5, and finally it is determined to be a non-grammatical sentence.

Similar to the sentence example (1), the sentence example (2) is "Yes" in step S1 and "N" in step S2.
o ", which is determined to be" Yes "in step S4,
In step S5, it is determined to be "Yes", and finally it is determined to be a 5W1H type interrogative sentence.

The sentence examples (3) and (4) are determined in the same manner as the sentence examples (1) and (2), respectively.

The example sentence (5) is "Y" in step S1.
es ", in step S2" Yes ", step S6
Is judged to be "No", and finally it is judged to be an affirmative sentence.

Similar to the sentence example (1), the sentence example (6) is "Yes" in step S1 and "Y" in step S2.
es "is determined, but" Yes "in step S6
It is determined that the question sentence is a Yes-No type interrogative sentence.

FIG. 2 shows an LFG (lexical-Functional G).
An example of the analysis result when the above sentence example (4) is subjected to syntactic analysis based on a grammar theory called rammar) is shown. The result of parsing based on LFG is c-structure.
It is called ure (constituent-structure) (described above). As can be seen from the figure, it is shown that the "who wrote" part is related to the "book" as a relative clause.

Further, FIG. 3 shows an example of the analysis result when the semantic analysis is further applied to the syntactic analysis result of the sentence example (4) based on the LFG. The result of semantic analysis based on LFG is f-structure (functi
onal-structure) is called (described above).

In the example shown in FIG. 3, f-structure
The INT mark is represented by the description "INT +" in e. The INT mark indicates that the function of requesting that the interrogative word "who" is an interrogative sentence extends. The INT mark is added to the verb phrase to which "who" is included, and also to the verb phrase to which the noun phrase to which the relative phrase belongs, when "who" is included in the relative phrase. f-structure
Is a sentence described in a structure in which the relative phrase is understood after the structural analysis and the semantic analysis are performed.
NT mark propagation processing is possible.

In the illustrated example, it can be seen that the INT mark added by "who" propagates to the upper part of the sentence, and the INT mark is added to the entire sentence.

As described above, when the grammatical theory of LFG is used, the present invention can be implemented more favorably by describing the grammatical rule of embedding the INT mark in f-structure.

According to the present invention, it is possible to analyze a question sentence with high accuracy. For example, when considering a dialogue system or a question answering system as the subsequent process, if the input by the user is determined to be a non-grammatical sentence, the correct meaning is returned to the user, or the 5W1H type or Yes.
-If it is determined that the question text is of the No type, a usage method of generating an answer according to the type can be considered.

According to the natural language processing system or the natural language method of the present invention, it is possible to determine with high accuracy whether or not the input sentence is a question sentence. You can answer questions by using it.

FIG. 7 schematically shows the functional configuration of the dialogue system 10 utilizing natural language processing according to the present invention. The dialogue system 10 includes a voice input unit 11, a voice recognition unit 12, a morpheme analysis unit 13, a syntax / semantic analysis unit 14, a context analysis unit 15, and an Azuma processing unit 16.

The voice input unit 11 inputs a voice spoken by a specific or unspecified speaker, and the voice recognition unit 12 recognizes the input voice and converts it into text format data. Of course, the input method of the input sentence is not limited to voice, and handwritten characters or text data may be input to the system 10.

The morpheme analysis unit 13 has a morpheme rule and a morpheme dictionary relating to a specific language such as Japanese, and divides the input sentence into morphemes, which are the smallest semantic units, and performs part-of-speech recognition processing.

The syntax / semantic analysis unit 14 has a dictionary such as a grammatical rule or a valence dictionary, and analyzes the phrase structure based on the grammatical rule or the like, and determines the meaning of words in a sentence or the semantic relationship between words. Based on this, we analyze the semantic structure that expresses the meaning conveyed by sentences. Then, as a result of the syntactic analysis, the input sentence is interrogated, past tense, and polite based on the c-structure that represents the phrase structure of the sentence composed of words and morphemes as a tree structure and the case structure of the subject and object. As a result of semantically and functionally analyzing sentences, f-structure
Is output.

In this embodiment, the syntax / semantic analysis unit 14 includes a question sentence determination unit 14A which determines with high accuracy whether or not the input sentence is a question sentence. This question sentence judgment section 1
In 4A, immediately after a word w belonging to a part-of-speech category W corresponding to an interrogative found in an input sentence, there exists a part-of-speech category P corresponding to a particle having a function of canceling the effect that the word w makes the entire sentence an interrogative sentence. Whether the input sentence is an interrogative sentence by determining whether or not the input sentence is the maximum possible phrase that is the maximum range of the function of requesting that the word w is the interrogative sentence in the input sentence. Is determined. That is, if it is determined that the maximum phrase covers the entire sentence, if the input sentence is a question sentence, it is determined that the input sentence is a 5W1H type question sentence, and if the determination result is negative, the input sentence is a non-grammar sentence. Judge as a target sentence.

If no word belonging to the part-of-speech category W is found in the input sentence,
When a word belonging to the part-of-speech category P is found immediately after the word w in the input sentence, and when it is determined that the maximum phrase does not reach the entire sentence, whether the entire input sentence is a question sentence or not Whether the input sentence is a Yes-No type interrogative sentence if the determination result is affirmative,
If the determination result is negative, it is determined that the input sentence is an affirmative sentence. Further, it is determined whether the input sentence is an interrogative sentence by determining whether the end of the input sentence ends with a final particle indicating a question or a symbol indicating an interrogation.

The context analysis unit 15 regards a sentence (discourse), which is a series of sentences, as a basic unit of analysis, and obtains a semantic cohesion between sentences to form a discourse structure.

The azuchi processing section 16 creates an azusa sentence for the input sentence using the result of the context analysis, and outputs it in a voice format, for example. In the present embodiment, the summation processing unit 16 includes an answer sentence creating unit 16A that generates an answer to the question sentence. The answer sentence creation unit 16A is provided by the question sentence determination unit 14
When it is determined by A that the input sentence is a Yes-No type interrogative sentence, an answer sentence based on the Yes-No format is generated. When the input sentence is determined to be a 5W1H type interrogative sentence, the response sentence creating unit 16A generates an answer sentence corresponding to the inquired matter.

[Supplement] The present invention has been described in detail with reference to the specific embodiments. However, it is obvious that those skilled in the art can modify or substitute the embodiments without departing from the scope of the present invention. That is, the present invention has been disclosed in the form of exemplification, and the contents of this specification should not be construed in a limited manner. In order to determine the gist of the present invention, the section of the claims described at the beginning should be taken into consideration.

[0075]

As described above in detail, according to the present invention,
It is possible to provide an excellent natural language processing system, natural language processing method, and computer program capable of suitably performing semantic analysis of Japanese syntax according to grammatical rules.

Further, according to the present invention, it is possible to accurately judge whether or not the inputted sentence is a question sentence without simply relying on the extraction of the question words. A language processing method and a computer program can be provided.

According to the present invention, whether the input sentence is a regular question sentence or non-sentence sentence according to the grammar, Yes
It is possible to provide an excellent natural language processing system and natural language processing method, and a computer program capable of determining whether the question sentence is a No type or a 5W1H type question sentence.

According to the present invention, it is possible to analyze a question sentence with high accuracy. For example, when considering a dialogue system or a question answering system as the subsequent process, if the input by the user is determined to be a non-grammatical sentence, the correct meaning is returned to the user, or the 5W1H type or Yes.
-If it is determined that the question text is of the No type, a usage method of generating an answer according to the type can be considered.

[Brief description of drawings]

FIG. 1 is a flowchart showing a processing procedure for determining whether an input sentence is a question sentence.

FIG. 2 is a diagram showing an example of an analysis result when a syntactic analysis is performed based on LFG.

FIG. 3 is a diagram showing an example of an analysis result when a semantic analysis is performed based on LFG.

FIG. 4 is a diagram schematically showing a configuration of a natural language processing system 1 based on LFG.

[Figure 5] Syntactically the input sentence "My daughter speaks English."
Fs obtained as a result of processing by the semantic analysis unit 1
It is the figure which showed structure.

[Fig.6] Syntactically the input sentence "My daughter speaks English."
Fs obtained as a result of processing by the semantic analysis unit 1
It is the figure which showed structure.

FIG. 7 is a diagram schematically showing a functional configuration of a dialogue system 10 using natural language processing according to the present invention.

[Explanation of symbols]

1. Natural language processing system 2 ... Morphological analysis unit 2A ... Morphological rules, 2B ... Morphological dictionary 3 ... Syntactic and Semantic Analysis Department 3A ... Grammar rule, 3B ... Bond valence dictionary 10 ... Dialog system 11 ... Voice input unit, 12 ... Voice recognition unit 13 ... Morphological analysis unit 14 ... Syntax / semantic analysis unit, 14A ... Question sentence determination unit 15 ... Context analysis section 16 ... Azuma processing section, 16A ... Answer sentence creating section

Claims (22)

[Claims] 1. A natural language processing system for determining whether an input sentence is an interrogative sentence, wherein it is determined whether the input sentence includes a word belonging to a part-of-speech category W corresponding to an interrogative sentence. The first means, and the word w immediately after the word w discovered by the first means.
A second means for determining whether or not there is a part-of-speech category P corresponding to a particle having a function of canceling the effect of making the entire sentence into a question sentence, and a natural language processing system. 2. A natural language processing system for determining whether or not an input sentence is a question sentence, and determines whether or not the input sentence includes a word belonging to a part-of-speech category W corresponding to an interrogative sentence. The first means, and the word w immediately after the word w discovered by the first means.
Second means for determining whether or not there is a part-of-speech category P corresponding to a particle having a function of canceling the effect of making the entire sentence into a question sentence, and a function for requesting that the word w in the input sentence is a question sentence And a third means for extracting the maximum possible phrase that is the maximum range of the input sentence, and whether or not the input sentence is a question sentence depending on whether the maximum phrase extracted by the third means covers the entire sentence. A natural language processing system comprising: a fourth means. 3. The natural language processing system according to claim 2, wherein the input sentence is a sentence described in a structure in which a relative phrase is understood. 4. The third means is a verb phrase to which the word w relates,
The natural language processing system according to claim 2, wherein, when the word w is included in the relative clause, the verb phrase related to the noun phrase related to the relative clause is extracted as the maximum possible phrase. 5. The method according to claim 4, further comprising: fifth means for determining whether or not the input sentence is a question sentence when the maximum phrase is determined to cover the entire sentence by the fourth means. Item 2. The natural language processing system according to item 2. 6. The fifth means determines that the input sentence is a 5W1H type interrogative sentence if the determination result is affirmative,
The natural language processing system according to claim 5, wherein if the determination result is negative, the input sentence is determined to be a non-grammatical sentence. 7. If the word belonging to the part-of-speech category W is not found in the input sentence by the first means, the second means is used.
Means that the part of speech category P immediately after the word w in the input sentence
When a word belonging to is found, or when it is determined by the fourth means that the maximum phrase does not extend to the entire sentence,
The natural language processing system according to claim 2, further comprising sixth means for determining whether or not the entire input sentence is a question sentence. 8. The sixth means determines that the input sentence is a Yes-No type interrogative sentence if the determination result is affirmative, and the input sentence is an affirmative sentence if the determination result is negative. The natural language processing system according to claim 7, wherein it is determined that there is. 9. The fifth means or the sixth means determines whether the input sentence is a question sentence or not, depending on whether the end of the input sentence ends with a final particle indicating a question or a symbol indicating a question.
The natural language processing system according to claim 5 or 6, characterized in that. 10. A natural language processing method for determining whether or not an input sentence is an interrogative sentence, wherein it is determined whether or not the input sentence includes a word belonging to a part-of-speech category W corresponding to the interrogative sentence. It is determined whether there is a part-of-speech category P corresponding to a particle having a function of canceling the effect that the word w makes the whole sentence a question sentence, immediately after the first step and the word w discovered by the first step. A second step of performing a natural language processing method. 11. A natural language processing method for determining whether an input sentence is an interrogative sentence, wherein it is determined whether the input sentence includes a word belonging to a part-of-speech category W corresponding to an interrogative sentence. It is determined whether there is a part-of-speech category P corresponding to a particle having a function of canceling the effect that the word w makes the whole sentence a question sentence, immediately after the first step and the word w discovered by the first step. And a third step of extracting a maximum possible phrase which is a maximum range covered by a function of requesting that the word w is a question sentence in the input sentence, and extracted by the third step. A fourth step of determining whether or not the input sentence is a question sentence depending on whether or not the maximum phrase covers the entire sentence, and a natural language processing method. 12. The natural language processing method according to claim 11, wherein the input sentence is a sentence described in a structure in which a relative phrase is understood. 13. In the third step, a verb phrase to which a word w relates and, if the word w is included in a relative clause, a verb phrase to which a noun phrase related to that relative phrase is extracted as a maximum possible phrase. The natural language processing method according to claim 11, wherein 14. The method according to claim 4, further comprising a fifth step of determining whether the input sentence is a question sentence when the maximum phrase is determined to cover the entire sentence in the fourth step. Item 12. The natural language processing method according to Item 11. 15. In the fifth step, if the determination result is affirmative, it is determined that the input sentence is a 5W1H type question sentence, and if the determination result is negative, the input sentence is a non-grammatical sentence. The natural language processing method according to claim 14, wherein it is determined that there is. 16. When a word belonging to a part-of-speech category W is not found in the input sentence in the first step,
When a word belonging to the part-of-speech category P is found immediately after the word w in the input sentence in the second step, or
The method further comprises a sixth step of determining whether or not the entire input sentence is a question sentence when it is determined that the maximum phrase does not reach the entire sentence in the fourth step. Item 12. The natural language processing method according to Item 11. 17. In the sixth step, if the determination result is affirmative, it is determined that the input sentence is a Yes-No type interrogative sentence, and if the determination result is negative, the input sentence is an affirmative sentence. The natural language processing method according to claim 16, wherein it is determined that there is. 18. In the fifth step or the sixth step, whether the sentence is a question sentence or not is determined depending on whether the end of the input sentence ends with a final particle indicating a question or a symbol indicating a question. The natural language processing method according to any one of claims 14 and 16, characterized in that. 19. A computer program written in a computer-readable format for executing natural language processing for determining whether an input sentence is an interrogative sentence on a computer system. A first step of determining whether or not a word belonging to a part-of-speech category W corresponding to a word is included, and an effect that the word w immediately after the word w discovered by the first step makes the entire sentence a question sentence And a second step of determining whether or not there is a part-of-speech category P corresponding to a particle having a function of canceling. 20. A computer program written in a computer-readable format so as to execute natural language processing for determining whether an input sentence is an interrogative sentence on a computer system. A first step of determining whether or not a word belonging to a part-of-speech category W corresponding to a word is included, and an effect that the word w immediately after the word w discovered by the first step makes the entire sentence a question sentence The second step of determining whether there is a part-of-speech category P corresponding to a particle having a function of canceling the particle, and the maximum possible range which is the maximum range covered by the function of requesting that the word w is a question sentence in the input sentence. A third step of extracting a phrase and a fourth step of determining whether or not the input sentence is a question sentence based on whether or not the maximum phrase extracted by the third step covers the entire sentence. A computer characterized by comprising:
program. 21. A question answering system for returning an answer to an inquiry, comprising: input means for inputting a minute from a user; and determining whether or not the input sentence includes a word belonging to a part-of-speech category W corresponding to an interrogative word. And a word w immediately after the word w discovered by the first means.
Second means for determining whether or not there is a part-of-speech category P corresponding to a particle having a function of canceling the effect of making the entire sentence into a question sentence, and a function for requesting that the word w in the input sentence is a question sentence And a third means for extracting the maximum possible phrase that is the maximum range of the input sentence, and whether or not the input sentence is a question sentence depending on whether the maximum phrase extracted by the third means covers the entire sentence. A question answering system comprising: a fourth means; and a fifth means for creating and outputting an answer sentence in response to determination of the question sentence. 22. A question answering method for returning an answer to an inquiry, comprising an input step of inputting a minute from a user, and determining whether or not the input sentence includes a word belonging to a part-of-speech category W corresponding to an interrogative word. And a word w immediately after the word w found by the first means.
The second step of determining whether there is a part-of-speech category P corresponding to a particle having a function of canceling the effect of making the entire sentence into a question sentence, and a function of requesting that the word w in the input sentence is a question sentence And a third step of extracting a maximum possible phrase that is the maximum range of the input sentence and determining whether the input sentence is an interrogative sentence by determining whether the maximum phrase extracted by the third means covers the entire sentence. A question answering method comprising: a fourth step; and a fifth step of creating and outputting an answer sentence in response to determination of the question sentence.