2002 Fiscal Year Final Research Report Summary
Automatic Text Summarization Using Robust Natural Language Processing Technologies
| Project/Area Number |
12680374
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Tokyo Institute of Technology |
|---|
Principal Investigator |
OKUMURA Manabu T.I.T., P&I Lab. , Associate Professor, 精密工学研究所, 助教授 (60214079)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
MOCHIZUKI Hajime Tokyo University of foreign Studies, Lecturer, 外国語学部, 講師 (70313707)
|
|---|
| Project Period (FY) |
2000 – 2002
|
| Keywords | Automatic Text Summarization / Coherence of a Text / Text Structure Analysis / Anaphora Resolution / Ellipsis Resolution |
|---|
| Research Abstract |
With the increasing number of on-line texts available, automatic text summarization has become one of the major research topics in the NLP community. In the field of the automatic summarization, the most major strategy is to extract important sentences from texts. And the main task of this approach is how to score the importance of sentences. This approach, producing extracts, that is, sets of extracted important sentences, can be considered to be easily realized, and therefore has been used as the major text summarization method for a long time.
However, as Paice pointed out, computer-produced extracts tend to suffer from the problem of 'lack of cohesion.' For example, the corresponding antecedents are not always included in an extract, even if anaphors are in the extract. This problem might cause the extracts to be difficult to read.
In this work, we first developed the system that analyzes the relation between sentences in a Japanese text and yields the text structure, the system that performs the ellipsis resolution in a Japanese text, and implemented the summarization system that uses the two discourse analysis systems.
|
Research Products
(6 results)
