JPH02236776A - Natural sentence meaning analysis processor - Google Patents

Abstract

PURPOSE:To absorb the error of a case postpositional particle, etc., by replacing an adjunct where there is the error with the adjunct specified by a meaning expression model and executing analysis even when there is the error in the adjunct such as the case postpositional particle in an input sentence. CONSTITUTION:In a morphenic analyzing means 2, the input sentence is divided into words and the respective words are classified to the adjuncts and independent words. In a meaning expression model collating means 3, at first, it is decided to which concept on a meaning expression model 1 the respective independent word belong and it is decided whether the word corresponding to the concept is connected by using the adjunct specified on the meaning expression model or not. Accordingly, the error of the adjunct is detected. When there is the error in the adjunct, the adjunct in the input sentence is replaced with the adjunct specified on the meaning expression model by a relation error correcting means 4 and meaning structure to the replaced sentence is generated by a meaning structure generating means 5. Thus, even when there is the error in the adjunct such as the case postpositional particle, etc., in the input sentence, the meaning structure of the sentence where the error is corrected can be outputted.

Description

[Detailed Description of the Invention] "Industrial Application Field" This invention analyzes the meaning of natural sentences read from a human-powered device and outputs a semantic structure that can be processed by a response processing device etc. connected at a later stage. Concerning semantic analysis processing equipment. "Conventional technology" Conventional semantic analysis processing devices perform morphological and syntactic analysis using dictionaries, grammar rules, and field-dependent knowledge, and generate semantic structures based on the generated dependency relationships and predefined rules. A semantic analysis processing device is known that performs semantic analysis processing by converting . ``Problems to be solved by the invention'' In conventional semantic analysis processing devices, if the input sentence contains a case particle error, the grammar rules are not satisfied, and the dependency relationship cannot be obtained and the analysis fails. However, since ordinary users do not know the grammatical rules defined by the semantic analysis processing device, they often input sentences containing case particle errors.

For this reason, when a dialogue processing system with a computer is configured using a conventional semantic analysis processing device, there is a problem that the smooth flow of dialogue is disrupted every time a sentence with a case particle error is input. .

The purpose of this invention is to replace errors in adjuncts such as case particles in an input sentence by replacing the erroneous adjuncts with adjuncts defined by a semantic representation model, thereby eliminating errors in case particles and other adjuncts. The purpose is to provide a semantic analysis processing device that can absorb information. ``Means and operations for solving the problem'' The semantic analysis processing device according to the present invention defines the concepts of the world that are subject to human power and the relationships between the concepts, and defines auxiliary words that indicate the relationships between the concepts. It has a semantic representation model, uses morphological analysis means to divide the input sentence into words, classifies them into attached words and independent words, and determines which concept on the semantic representation model the words classified as independent words belong to. and detecting an error in the attached word by determining whether the word corresponding to the concept is connected using the attached word defined on the semantic expression model by the semantic expression model matching means, If there is an error in the attached word, the attached word in the input sentence is replaced by the attached word specified on the semantic expression model by the dependency error correction means, and the semantic structure for the replaced sentence is generated by the semantic structure generation means. By doing this, even if there are errors in attached words such as case particles in the input sentence, the semantic structure of the sentence with the errors corrected can be output. "Embodiment" Figure 1 is a block diagram showing an embodiment of this invention. In FIG. 1, 1 is a semantic expression model, 2 is a morphological analysis means, 3 is a semantic expression model matching means, 4 is a dependency error correction means, and 5 is a semantic structure generation means. FIG. 2 shows an example of the structure of a semantic expression model. In Figure 2, the object class, event class, and value class represent concepts, and the arrows represent the relationships between these concepts. The operation of the present invention will be explained using FIG.

The input sentence is input to the morphological analysis means 2. The morphological analysis means 2 divides the input sentence into words, and divides each word into an adjunct word and an independent word. For example, [I want to stay at Fujiya Ryokan.]
” is input. "Fujiya Ryokan" and "stay" are extracted as independent words.

The semantic expression model matching means 3 first determines which concept on the semantic expression model 1 each independent word belongs to. “Fujiya Ryokan” belongs to the concept of “inn name゜゛゛,” and “staying” belongs to the concept of “staying゜゛.” In semantic expression model 1 in Figure 2, superordinate concepts are defined by isa relationships and name relationships. ．For example, as a superordinate concept of “inn name”
Because accommodation is stipulated, “Fujiya Ryokan” is an “inn”.
It belongs to the concept of In the semantic representation model l, ゜゜ inn
The relationship between "Fujiya Ryokan" and "stay" are connected by the adjunct "ni", whereas in the input sentence "Fujiya Ryokan" and "stay" are connected by "he". Therefore, "he" is determined to be an error, and this result is sent to the 4th stage for error correction. The dependency error correction means 4 replaces the attached word determined to be an error with the attached word defined on the semantic expression model l.

That is, "he" is replaced with "ni", and the sentence "I want to stay at Fujiya Inn." is generated and sent to the semantic structure generation means 5. In the semantic structure generation means 5, "I want to stay at Fujiya Ryokan.
A semantic structure is generated for the sentence ``. The sentence corrected by the dependency error correction means 4 can output a correct semantic structure as long as it satisfies the relationship defined by the semantic expression model l. "Effects of the Invention" As explained above, according to the present invention, even if there is an error in an adjunct such as a case particle in an input sentence, the incorrect adjunct is replaced with an adjunct defined by the semantic representation model and analyzed. Therefore, it is possible to realize a dialogue processing system that can maintain a smooth flow of dialogue even if a sentence with an erroneous adjunct is input.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, and FIG. 2 is a diagram showing an example of the configuration of a semantic expression model. Patent applicant Nippon Telegraph and Telephone Corporation

Claims (1)

[Claims] (1) In a natural sentence semantic analysis processing device having means for inputting an input sentence and means for outputting a result of semantic analysis of the input sentence, defining concepts of the world to be input and relationships between the concepts; a meaning expression model that defines attached words that indicate relationships between concepts; a morphological analysis means that divides the input sentence into words and classifies them into attached words and independent words; The meaning of determining which concept on the expression model the word belongs to, and determining whether the adjunct word in the input sentence that connects the words is an adjunct word that relates the concepts on the semantic expression model. expression model matching means, and when the judgment result of the semantic expression model matching means is negative, replacing the attached word in the input sentence with the attached word defined on the semantic expression model between the concepts; a dependency error correction unit that generates a sentence; and a semantic structure generation unit that generates a semantic structure for a sentence that is not determined to be negative by the semantic representation model matching unit and a sentence that is corrected by the dependency error correction unit. A natural sentence semantic analysis processing device comprising: