Learning Robust Word Representations for Natural Language Processing

2013 Fiscal Year Final Research Report

• Project/Area Number: 24800041
• 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: DUH Kevin 奈良先端科学技術大学院大学, 情報科学研究科, 助教 (80637322)
• Project Period (FY): 2012-08-31 – 2014-03-31
• Keywords: 自然言語処理 / 機械学習

Research Abstract

Language is a highly productive phenomenon; new words and expressions are constantly being invented. Current Natural Language Processing techniques have difficulty handling new words, so their performance on real-world text suffers. To address this, we develop robust models of word semantics, focusing on Deep Learning methods for learning vector word representations. Furthermore, we show improvements in parsing and translation performance using systems that incorporate these word representations.

Research Products

• [Presentation] Modeling and Learning Semantic Co-Compositionality through Prototype Projection and Neural Networks (2013)
  • Author(s): M Tsubaki, K Duh, M Shimbo, and Y Matsumoto
  • Organizer: Proceedings of the Conference on Empirical Methods in Natural Language Processing (EMNLP)
  • Place of Presentation: Seattle, USA.
  • Year and Date: 2013-10-18
• [Presentation] Adaptation Data Selection using Neural Language Models : Experiments in Machine Translation (2013)
  • Author(s): K Duh, G Neubig, K Sudoh, and H Tsukada
  • Organizer: Proceedings of the Annual Meeting of the Association for Computational Linguistics (ACL)
  • Place of Presentation: Sofia, Bulgaria
  • Year and Date: 2013-08-06
• [Presentation] An Empirical Investigation of Word Representations for Parsing the Web (2013)
  • Author(s): S Hisamoto, K Duh, and Y Matsumoto
  • Organizer: 言語処理学会第19回年次大会発表論文集
  • Place of Presentation: 名古屋大学
  • Year and Date: 2013-03-14