1993 Fiscal Year Final Research Report Summary
Study of Efficient Parsing Methods Based on Japanese Phrase Structure Grammar
| Project/Area Number |
03452169
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
情報工学
|
|---|
| Research Institution | Osaka University |
|---|
Principal Investigator |
GUNJI Takao Osaka University, Graduate School of Language and Culture, Associate Professor, 大学院・言語文化研究科, 助教授 (10158892)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
MIKI Kunihiro Sugiyama Women's University Faculty of Life Science, Associate Professor, 生活科学部, 助教授 (80174001)
|
|---|
| Project Period (FY) |
1991 – 1993
|
| Keywords | Phrase Structure Gramr / Parsing / Constraint-based Gramr / Logic Programming |
|---|
| Research Abstract |
In this study, we have attempted to develop a gramr description system suited for computer processing of natural language. Our focus was on mechanisms inherent in grammatical theory, in particular, on how constraints are described. In order to achieve this goal, we have developed an efficient gramr description formalism with which to implement grammatical theory. With this formalism, we have developed a prototype of a parsing system and implemented the core part of Japanese Phrase Structure Grammar on it.
In 1991 and 1992, our focus was on the genaral method of describing constraints associated with grammatical generalizations. We also paid attention to natural and efficient ways for implementing linguistically-required grammatical constraints in a computational environment. In 1992, follwing these results, we have developed a parsing mechanism and a grammatical description formalism employed by the parser as separate modules and tried to make the crossreferencing procedure explicit.
In 1992 and 1993, we first transformed the existing top-down parsing algorithm into a bottomup program and developed some tools for minimizing the discrepance between grammatical description and its implementation.
As the result, we have found out that multi-layred feature representation for sentence structures we have proposed in this research not only make it possible to uniformly treat morphological representation and semantic representation but also make it feasible to develop a parsing system based on such representation.
Even though we have first attempted to introduce the concept of continuous cost into our inventory of constraints, we have not arrived at a satisfactory result on the way of introducing such continuous concept in the discrete system of logic programming. We hope to handle this problem in a separate research project in the near future.
|
