2002 Fiscal Year Final Research Report Summary
INTERPRETATION PROCESS OF JAPANESE REFERRING EXPRESSIONS : FROM THEORY TO APPLICATION
| Project/Area Number |
12680396
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | INTERNATIONAL CHRISTIAN UNIVERSITY |
|---|
Principal Investigator |
MATSUI Tomoko INTERNATIONAL CHRISTIAN UNIVERSITY, ASSOCIATE PROFESSOR, 教養学部, 準教授 (20296792)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
MATSUMOTO Yuji NARA INSTITUTE OF SCIENCE AND TECHNOLOGY, GRADUATE SCHOOL OF INFORMATION SCIENCE, 情報科学研究科, 教授 (10211575)
|
|---|
| Project Period (FY) |
2000 – 2002
|
| Keywords | ZERO-PRONOUNS / REFERENCE RESOLUTION / RELEVANCE THEORY / CENTERING THEORY / PRAGMATICS |
|---|
| Research Abstract |
Interpretation of anaphoric reference is highly context-dependent, and Japanese zero-pronouns are the clearest examples of that. The present study has two main goals. The first is to compare and assess several leading theories of reference assignment, in order to propose a theory of our own which explains the interpretation mechanism of Japanese zero-pronouns. The second is to verify the theory by applying it to create a workable computer program for pronoun resolution.
The following are is the main achievements of the study:
(1)A theoretical model of reference resolution, based on relevance theory (Sperber & Wilson 1995) has been proposed. The key mechanisms that are involved in interpretation of Japanese zero-pronouns have been identified, such as the balance between 'contextual effects' and 'processing effort' and interaction between 'implicature' (i.e. expectation of specific contextual effects) and 'explicature' (i.e. linguistic information).
(2)A new experiment designed to test on-line interpretation of anaphoric reference has started. This experimental study will be completed within 2 years from now.
(3)A machine learning of model for reference resolution has been proposed and tested. The model incorporate linguistically motivated contextual clues (centering theory). It was applied to resolve Japanese zero-pronouns and outperformed earlier machine learning models.
|
Research Products
(14 results)
