A Study of Practical Spoken Language Processing

2016 Fiscal Year Final Research Report

• Project/Area Number: 15H06408
• Research Category: Grant-in-Aid for Research Activity Start-up
• Allocation Type: Single-year Grants
• Research Field: Intelligent informatics
• Research Institution: Nara Institute of Science and Technology
• Principal Investigator: Koichiro Yoshino 奈良先端科学技術大学院大学, 情報科学研究科, 助教 (70760148)
• Project Period (FY): 2015-08-28 – 2017-03-31
• Keywords: 述語項構造 / 対話システム

Outline of Final Research Achievements

We developed a method of semantic analysis, especially for predicate argument structure on spoken language. The proposed method considers backward reference, a characteristic of spoken language to robustly process spoken languages. We developed a predicate argument structure analyzer that does not use dependency information of words to verify the influence of lack of dependency information, which is difficult to extract from spoken sentences. We used pointwise method to build the system to analyze sentences, even if it is unfinished sentence, by introducing backward reference label. This architecture make it possible to predict not only structure of spoken contents by the user but also existences of unseen future content words. We also developed a dialogue corpus to confirm the phenomena of backward reference.

Free Research Field

音声対話