Composite Probabilistic Grammar Incorprating Word Dependency Constraint
| Project/Area Number |
09680370
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | KYUSHU UNIVERSITY |
|---|
Principal Investigator |
HITAKA Toru Kyushu Univ.Graduate School of Information & Electrical Engineering Professor, 大学院・システム情報科学研究科, 教授 (30037931)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
ICHIMARU Natsuki Kyushu Univ.Graduate School of Information & Elec-trical Engineering Research As, 大学院・システム情報科学研究科, 助手 (80274497)
TOMIURA Yoich Kyushu Univ.Graduate School of Information & Elec-trical Engineering Associate P, 大学院・システム情報科学研究科, 助教授 (10217523)
|
|---|
| Project Period (FY) |
1997 – 1998
|
| Project Status |
Completed (Fiscal Year 1998)
|
| Budget Amount *help |
¥3,900,000 (Direct Cost: ¥3,900,000)
Fiscal Year 1998: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 1997: ¥3,100,000 (Direct Cost: ¥3,100,000)
|
|---|
| Keywords | probabilistic Grammar / Composite Probabilistic Grammar / Word Dependency Constraint / Constraint Grammar / 確率パラメタ推定式 / 係り受け関係 |
|---|
| Research Abstract |
Obectives of this research are to develop (1) a method to systematically incorporate word dependency constraint into a phrase structure grammar, and (2) a composite probabilistic grammar model with higher likelifood which offers a method for better use of a probabilistic grammar. Those two methods are very useful for avoiding syntactic ambiguity in natural language processing. Obtained main results are as follows.
(1) A method to incorporate word-dependency constraint into a phrase structure grammar is proposed. Using triplex of category, head word and function word as refined categories, we proved that word dependency constraint, which is very efficient to decrease syntactic ambiguity, is systematically incorporated into a new phrase structure grammar.
(2) The probabilistic context free grammar has a set of parameters. Using m sets of parameters and distributing coefficients to the m sets, a probabilistic context grammar is transfered into a composite probabilistic grammar. The composite probabilistic grammar is proved to have higher likelifood than original probabilistic grammar, and formula of m sets of parameters and distributing coefficients are driven by Baum theory.
|
Report
(3 results)
Research Products
(24 results)
