Large-scale text data analysis using hashing techniques

• Project/Area Number: 26730126
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Intelligent informatics
• Research Institution: NTT Communication Science Laboratories
• Principal Investigator: Hayashi Katsuhiko 日本電信電話株式会社NTTコミュニケーション科学基礎研究所, 協創情報研究部, 研究員 (50725794)
• Project Period (FY): 2014-04-01 – 2017-03-31
• Project Status: Completed (Fiscal Year 2016)
• Budget Amount: ¥2,470,000 (Direct Cost: ¥1,900,000、Indirect Cost: ¥570,000); Fiscal Year 2016: ¥390,000 (Direct Cost: ¥300,000、Indirect Cost: ¥90,000); Fiscal Year 2015: ¥260,000 (Direct Cost: ¥200,000、Indirect Cost: ¥60,000); Fiscal Year 2014: ¥1,820,000 (Direct Cost: ¥1,400,000、Indirect Cost: ¥420,000)
• Keywords: 談話構造解析 / 省略補完 / 行列分解 / ハッシュ法 / 低ランク近似 / 分枝限定法 / 線形時間言語解析 / 音声言語データ解析 / 自然言語処理 / 談話解析 / 修辞構造解析

Outline of Final Research Achievements

I investigated hashing and matrix factorization techniques to efficiently analyze large-scale text data in various domains. First, I proposed a fast and accurate parsing algorithm for discourse tree structure analysis of English newswire texts. I also presented a text summarization method using discourse trees, and achieved an improvement in text summarization accuracy. Second, I proposed a method to automatically detect and insert missing elements in English and Japanese speech/newswire texts. Finally, I proposed a knowledge (word thesaurus) embedding method for fast word similarity computation. In future, I will apply these methods to such more advanced NLP applications as machine translation and question answering.

Research Products

• [Presentation] On the Equivalence of Holographic and Complex Embeddings for Link Prediction (2017)
  • Author(s): Katsuhiko Hayashi, Masashi Shimbo
  • Organizer: The 55th Annual Meeting of the Association for Computational Linguistics
  • Place of Presentation: バンクーバー
  • Year and Date: 2017-07-31
  • Int'l Joint Research
• [Presentation] K-best Iterative Viterbi Parsing (2017)
  • Author(s): Katsuhiko Hayashi, Masaaki Nagata
  • Organizer: The 15th Conference of the European Chapter of the Association for Computational Linguistics
  • Place of Presentation: バレンシア
  • Year and Date: 2017-04-05
  • Int'l Joint Research
• [Presentation] 知識グラフの埋め込みとその応用 (2017)
  • Author(s): 林克彦
  • Organizer: 千葉工業大学 ステアラボ人工知能セミナー
  • Place of Presentation: 東京
  • Invited
• [Presentation] Empirical comparison of dependency conversions for RST discourse trees (2016)
  • Author(s): Katsuhiko Hayashi, Tsutomu Hirao, Masaaki Nagata
  • Organizer: The 17th Annual Meeting of the Special Interest Group on Discourse and Dialogue
  • Place of Presentation: ロサンゼルス
  • Year and Date: 2016-09-13
  • Int'l Joint Research
• [Presentation] Empty element recovery by spinal parser operations (2016)
  • Author(s): Katsuhiko Hayashi
  • Organizer: The 54th Annual Meeting of the Association for Computational Linguistics
  • Place of Presentation: ベルリン
  • Year and Date: 2016-08-08
  • Int'l Joint Research
• [Presentation] 修辞構造木から自動変換した談話依存構造木の性質について (2015)
  • Author(s): 林克彦
  • Organizer: 言語処理学会年次大会
  • Place of Presentation: 京都大学(京都府京都市)
  • Year and Date: 2015-03-16 – 2015-03-21
• [Patent(Industrial Property Rights)] 単語学習装置、単語学習方法及び単語学習プログラム (2017)
  • Inventor(s): 林克彦、新保仁、永田昌明
  • Industrial Property Rights Holder: 林克彦、新保仁、永田昌明
  • Industrial Property Rights Type: 特許
  • Filing Date: 2017-03-02