2000 Fiscal Year Final Research Report Summary
Methods of Resolving Word Sense and Syntactic Ambiguities Using Information on Coprora
| Project/Area Number |
11680381
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | The University of Electro-Communications |
|---|
Principal Investigator |
FURUGORI Teiji Univ. of Electro-Communications, Dept. of Computer Science, Professor, 電気通信学部, 教授 (80114932)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
WU Haodong Dokkyo University, Department of Foreign Languages, Associate Professor, 外国語学部, 助教授 (50293121)
|
|---|
| Project Period (FY) |
1999 – 2000
|
| Keywords | Ambiguity / Natural Language / Disambiguation / Co-occurrence / Similarity / Corpus |
|---|
| Research Abstract |
This research proposes methods for resolving ambiguities appearing on word meanings and syntactic structures. We put them in algorithms and programs and test their effectiveness experimentally.
The languages studied are Japanese, chinese, and English. For Japanese, we tried to resolve ambiguities that appear in the noun phrases in which nouns are connected by a number of no's. For Chinese, an attempt was made to segment words correctly by resolving overlapping ambiguities. For English, we developed a method for word sense disambiguation. The methods used are all similar. We base our methods on word co-occurrence relations and word similarity measures. These are likely means we humans use to reslove ambiguities in natural language : we know that bank in I go to the bank to get money means a commercial institution rather than a river side because it appears with money. We derive such knowledge from various corpora and measures co-occurrence relaions and similarities statistically. The results we got from our experiments are all fevorable when compared with other studies.
|
