JPH07121543A - Convensational sentence analyzing device - Google Patents

Abstract

PURPOSE:To analyze a conversational sentence similarly to vocabularies for written words. CONSTITUTION:A conversational sentence analyzing device 100 consists of an analytical dictionary 110, a syntax analyzing rule 120 and a syntax semantic analyzing part 130. A written word dictionary 111 and a conversational sentence dictionary 112 are prepared as the dictionary 110 and infixes (e.g. 'ano', 'etto') and abbreviated words (e.g. 'ja') are previously registered in the dictionary 112. The analyzing part 130 judges parts of speech for respective words in an inputted conversational sentence by the use of both the dictionaries 111, 112 and repeatedly executes processing for applying the rule 120 to the judged words to find out tree structure for the conversational sentence. The meaning expression of the conversational sentence is determined by the tree structure and outputted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a conversational sentence analysis device, for example, a character string that is used by a person in daily life through a voice recognition device, a character recognition device, a keyboard of a computer terminal, or a combination thereof. It is related to the technology of inputting into a computer as, and processing by the computer what the character string means.

[0002]

2. Description of the Related Art Conventional natural language processing techniques are intended mainly for written words. Words that frequently appear in spoken words, such as "Ah" and "Etto", and "(- Inserted words such as phrases inserted at breaks in phrases such as "ha"), and "sore" used in place of "there" and "going" used in place of "going" It is not suitable for analyzing sentences including contracted words such as. In order to handle synonym sentences, in the preprocessing, it was necessary to remove inserted words or convert contracted words into vocabulary of written language, and then perform syntactic and semantic analysis.

[0003]

A method of performing syntactic and semantic analysis after removing an insertion word or a contraction word peculiar to a conversational sentence in a preprocessing or converting it into a vocabulary of a written word is as follows. There are drawbacks.

When the syntactic and semantic analysis is performed after the inserted word is found and removed in the preprocessing, it looks like "is".
Words used in addition to the inserted words are also removed. For example, in the sentence "I think about that, I will think about it", you can remove "
In the sentence "The manuscript deadline is 31st," it cannot be removed because "ne" has the function of requesting confirmation. In the conventional method, the case where it is used other than such an insertion word is also removed.

There is also a problem in converting a contracted word into a written word. For example, the word "ja" is used as an abbreviation for "wa", but not all "ja" in conversational sentences are replaced with "wa".

An object of the present invention is to enable the analysis of conversational sentences by treating them in the same way as the vocabulary of written words without using a special procedure for treating insertion words and contracted words peculiar to conversational sentences. It is in.

[0007]

In order to achieve the above object, the present invention comprises a dictionary for written words, a syntactic analysis rule, and a syntactic and semantic analysis unit for performing syntactic and semantic analysis of an input sentence using these. In the sentence analysis device, a conversational sentence dictionary is added to the written word dictionary, and syntactic and semantic information such as reduced expressions and redundant expressions peculiar to the conversational sentence is registered in the conversational sentence dictionary, and the syntactic and semantic analysis unit is described. Is characterized in that, in addition to the written word dictionary, the conversation sentence dictionary is used to perform syntactic and semantic analysis of the input conversation sentence based on the syntactic analysis rule.

[0008]

[Operation] Inserted words and contracted words are registered in the conversation sentence dictionary. For example, an insertion word such as "an" or "umu" modifies a verb phrase, but is registered as a modifier that does not affect the meaning. In addition, a word inserted after a particle such as "is" is registered as a particle that follows the particle phrase and does not affect its meaning. Further, the contracted word “ja”, which replaces “”, is registered as a word having the same function as “”. In this way, by registering the words specific to the conversational sentence in the form of being added to the dictionary for the written language, the syntactic and semantic analysis of the conversational sentence can be performed using the same rules and the syntactic and semantic analysis method as the written language.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention. In the figure, the conversational sentence analysis device 100 includes an analysis dictionary 110, a syntactic analysis rule 120, and a syntactic and semantic analysis unit 130. In the dictionary 110, a conversation sentence dictionary 112 is prepared in addition to the written word dictionary 111.
The insertion words and contracted words peculiar to the conversational sentence are registered in 2. When a conversational sentence is input, the syntactic and semantic analysis unit 130 refers to the conversational sentence dictionary 112 in addition to the written word dictionary 110, analyzes the syntactical and semantics of the input conversational sentence according to the syntactic rule 120, and analyzes the input conversational sentence. Output a semantic expression.

FIG. 2 shows a block diagram when used in a natural language processing system, as an example of a mode of use of the conversational sentence analysis apparatus. In the figure, the expression (speech) of the conversational sentence to be processed is input to the voice recognition device 10, subjected to recognition processing, converted into a character string, and sent to the conversational sentence analysis device 20. The conversational sentence analysis device 20 has the configuration shown in FIG.
In 0, by referring to the dictionary 112 for conversational sentences in addition to the dictionary 111 for written sentences, the syntactic and semantic analysis of the character string of the input conversational sentence is performed using the same syntax rule 120 as the written language, The meaning expression is output and sent to the response generation device 30. The response generation device 30 plans an appropriate response from the semantic representation of the conversational sentence, and outputs the response by voice using the voice synthesis device 40.

[0012] Next, as an example, "I will think about it, I will consider it until now" (hereinafter referred to as example sentence 1) and "The manuscript deadline is 31 days" (hereinafter referred to as example sentence 2). The syntactic and semantic analysis processing of a conversation sentence is shown.

The written sentence dictionary 111, the conversational sentence dictionary 112, and the syntax rules 120 of the conversational sentence analysis device 100 of FIG.
Is assumed to be specifically as shown in FIGS. 3, 4, and 5, respectively. Here, each line in FIG. 5 corresponds to one syntax rule. For example, in the first syntax rule "verb phrase ← noun phrase verb phrase", a sequence of noun phrases and verb phrases constitutes a verb phrase. Means

The syntactic and semantic analysis unit 1 of the conversational sentence analysis apparatus 100.
In 30, the written word dictionary 111 and the conversational sentence dictionary 1
12, the part of speech of each word of the input conversational sentence is determined, and the process of applying the syntax rule 120 to this is repeatedly executed to obtain the tree structure of the input conversational sentence. From the obtained tree structure, Determine the semantic expression of the conversation sentence.

In the case of example sentence 1, the tree structure is obtained as shown in FIG. Here, when focusing on "about that," from the dictionary for written words in FIG. 3, "that" is a noun (pronoun), "about" is a particle, and the third syntactic rule "noun" in FIG. From "phrase ← noun particle", the part "about it" becomes a noun phrase. Next "is" is "is" + "ne"
However, it is ambiguous whether it is an interjection "is". Therefore, assuming that "is" + "ne" is a decision word in the written word dictionary of FIG. 3, the syntactic rule of FIG. 5 does not have a sequence of noun phrases and decision words. On the other hand, if it is "is", this is registered as an interjection particle in the conversational sentence dictionary in FIG. 4, and in FIG. 5, there is a syntax rule of "noun phrase ← noun phrase interpositional particle" (6 ). As a result, it can be seen that “” about “that” was an interjection particle. The analysis of the “I will think about this time” part is the same as before. After all, the semantic representation of the example sentence 1 becomes as shown in FIG. 7 from the tree structure of FIG.

In the case of example sentence 2, the tree structure is as shown in FIG. 8 and the semantic representation is as shown in FIG. In this case, according to the syntax rule “determinant phrase ← noun determiner” (10th) and “sentence ← sentence final particle” (12th) in FIG. 5, “is” is “is” (determinant) + “ne”. It turns out that it is the (ending point).

[0017]

By using the conversation analysis device of the present invention, it is possible to perform a syntactic and semantic analysis of a conversation sentence which cannot be handled by the conventional natural language processing technique for written words. Further, according to the present invention, it is possible to analyze a conversational sentence without performing preprocessing. In the preprocessing, before performing the syntactic and semantic analysis of the sentence,
Since the sentence is converted, there is a possibility of making an incorrect conversion, but the present invention is more accurate because it recognizes a word peculiar to a conversational sentence at the same time as the syntactic and semantic analysis. By using the conversation analysis device of the present invention, the natural language processing system by voice input can handle more kinds of sentences.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of a conversational sentence analysis device of the present invention.

FIG. 2 is a diagram showing an example of a usage mode of the conversational sentence analysis device of the present invention.

FIG. 3 is a diagram showing an example of a written word dictionary.

FIG. 4 is a diagram showing an example of a conversational sentence dictionary.

FIG. 5 is a diagram showing an example of syntax rules.

6 is a tree structure of example sentence 1. FIG.

7 is a semantic representation of example sentence 1. FIG.

8 is a tree structure of example sentence 2. FIG.

9 is a semantic representation of example sentence 2. FIG.

[Explanation of symbols]

 100 Conversational sentence analysis device 110 Analysis dictionary 110 111 Written language dictionary 112 Conversational sentence dictionary 120 Syntax rules 130 Syntax meaning analysis unit

Claims (1)

[Claims] 1. A sentence analysis device comprising a written word dictionary, a syntactic analysis rule, and a syntactic and semantic analysis unit for performing syntactic and semantic analysis of an input sentence using these, and a conversational sentence dictionary is added to the written word dictionary. Then, the syntactic and semantic information such as a contracted expression and a redundant expression peculiar to the conversational sentence is registered in the conversational sentence dictionary, and the syntactic and semantic analysis unit uses the conversational sentence dictionary in addition to the written word dictionary. A conversational sentence analysis device for performing a syntactic and semantic analysis of an input conversational sentence based on the syntactic analysis rule.