1992 Fiscal Year Final Research Report Summary
Lexical Functional Model for Discourse Understanding and its Application to Natural Language Processing
| Project/Area Number |
02452167
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
情報工学
|
|---|
| Research Institution | Kyushu Institute of Technology |
|---|
Principal Investigator |
NOMURA Hirosato Kyushu Institute of Technology, Professor, 情報工学部, 教授 (30208392)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
TSUDA Ichiro Kyushu Institute of Technology, Assistant Professor, 情報工学部, 助教授 (10207384)
GOTO Mariko Kyushu Institute of Technology, Lecturer, 情報工学部, 講師 (20189773)
SUGIMOTO Takeshi Kyushu Institute of Technology, Lecturer, 情報工学部, 講師 (70196749)
NAGAI Hidetoshi Kyushu Institute of Technology, Assistant, 情報工学部, 助手 (60237485)
|
|---|
| Project Period (FY) |
1990 – 1992
|
| Keywords | computational linguistics / linguistics / natural language processing / discourse understanding / lexical functional model |
|---|
| Research Abstract |
To study on computational discourse understanding, we take note of linguistic function of word. We analyze the function both linguistically and computationally. On the basis of the analysis, we design a discourse understanding model, and develop an experimental system based on the model.
The followings are some of the research results.
1. We analyze one of the sentence endings "(te-simau)". "(te-simau)" has "completion" type and "realization" type of aspect. Sometimes the "realization" type of "(te-simau)" has a kind of modality. Based on the analysis of the "realization" type of "(te-simau)", we show that the type represents "the occurrence of situation which can not be controlled by the speaker". And we show the position of "(te-simau)" on aspect and modality.
2. We propose a controlled linguistic model for law sentences. The linguistic model is investigated from the analysis of the real law sentences. The law sentences are analyzed in view of their characteristics of sentence endings which are controlled both syntactically and semantically. Based on the classification of "what the law sentences govern", we investigate patterns of logical structure of law sentences. We discuss a linguistic model by which we can specify those patterns.
3. We propose a document summarization method which produces a chart/table as a summary of text. The whole text is represented as an integration of extended case structures of each sentence. The integration becomes a complex semantic network. The semantic network is reduced by applying "contribution degree" which specifies the significance of each concept represented by a case element. The reduced network is converted into a chart/table representation consisting of "cart elements".
|
