A model for natural language understanding based on spatio-temporal continuity and its application to natural language processing
| Project/Area Number |
15K02475
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Research Field |
Linguistics
|
|---|
| Research Institution | Shizuoka University |
|---|
Principal Investigator |
KONDO Makoto 静岡大学, 情報学部, 教授 (30225627)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
小西 達裕 静岡大学, 情報学部, 教授 (30234800)
小暮 悟 静岡大学, 情報学部, 准教授 (40359758)
野口 靖浩 静岡大学, 情報学部, 講師 (50536919)
|
|---|
| Project Period (FY) |
2015-04-01 – 2019-03-31
|
| Project Status |
Completed (Fiscal Year 2018)
|
| Budget Amount *help |
¥4,030,000 (Direct Cost: ¥3,100,000、Indirect Cost: ¥930,000)
Fiscal Year 2018: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2017: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2016: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2015: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
|
|---|
| Keywords | 文脈参照 / 構文意味理解 / 自然言語処理 / 語用論 |
|---|
| Outline of Final Research Achievements |
In a natural language dialog, the speaker makes utterances by referring to the preceding context, and the hearer also refers to the context in understanding the utterances. As a result, not every utterance explicitly includes information on the time, place, etc. of the event mentioned in the utterance. When such information is recoverable from the context, they are usually omitted in an utterance. This research focuses on understanding the time and place of an event in an utterance. We have developed a model that enables spatio-temporal interpretation referring to the context and implemented a meaning-understanding component for a natural language processing system.
|
| Academic Significance and Societal Importance of the Research Achievements |
近年、自然言語処理分野で大きな成果を挙げている大規模コーパスに基づく統計的手法や機械学習（特に深層学習）を用いた手法は、原理的に処理対象となる文1つ1つの意味を理解したうえで処理を行っているわけではなく、入力から確率的に出力を求める手法であると言える。本研究では処理対象となる個々の文の意味を意味表現というかたちで明示的に表し、1つ1つの文の意味の理解に基づいて発話文（入力文）と文脈との関係を明示的に表現するモデルを開発し実装した。その意味において、本研究は確率的手法とはまったく異なる独自の手法を提案している。
|
Report
(5 results)
Research Products
(6 results)
