Self-explainable and fast-to-train example-based machine translation using neural networks

• Project/Area Number: 18K11447
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 61030:Intelligent informatics-related
• Research Institution: Waseda University
• Principal Investigator: LEPAGE YVES 早稲田大学, 理工学術院(情報生産システム研究科・センター), 教授 (70573608)
• Project Period (FY): 2018-04-01 – 2021-03-31
• Project Status: Completed (Fiscal Year 2020)
• Budget Amount: ¥4,030,000 (Direct Cost: ¥3,100,000、Indirect Cost: ¥930,000); Fiscal Year 2020: ¥780,000 (Direct Cost: ¥600,000、Indirect Cost: ¥180,000); Fiscal Year 2019: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000); Fiscal Year 2018: ¥1,690,000 (Direct Cost: ¥1,300,000、Indirect Cost: ¥390,000)
• Keywords: 自然言語処理 / 機械翻訳 / 用例手法 / ニューラルと統計手法 / analogy / machine translation / natural language / case-based reasoning / explainable AI

Outline of Final Research Achievements

This research introduced self-explanation in example-based machine translation (EBMT) by analogy. It is thus positioned in explainable artificial intelligence (XAI). Self-explanation was implemented by tracing the analogies verified or solved during translation. The direct and indirect approaches to EBMT by analogy were merged in system that uses an original neural network. It was studied how to retrieve sentences that cover a given sentence semantically and formally was built. It was studied how dense corpora are relative to analogies. Datasets of analogies between sentences were released.

Academic Significance and Societal Importance of the Research Achievements

統計的機械翻訳（SMT）の手法である文部分アライメントと、ニューラル自然言語処理（NMT）の手法である単語や文のベクトル表現を用いて、類推関係方程式の解を求める手法を改善した。類推関係に基づいた用例機械翻訳の直接的なアプローチと間接的なアプローチを融合し、独自のニューラルネットワークを用いたシステムを構築した。入力は、単語のベクトル表現に基づく、単言語または対言語のソフトアライメントです。

Research Products

• [Journal Article] Vector-to-sequence models for sentence analogies (2020)
  • Author(s): L. Wang and Y. Lepage
  • Journal Title: In IEEE, editor, Proceedings of the 2020 International Conference on Advanced Computer Science and Information Systems (ICACSIS 2020) Volume: ICACSIS 2020 Pages: 441-446
  • DOI: 10.1109/icacsis51025.2020.9263191
  • Peer Reviewed / Open Access
• [Journal Article] Reseaux de neurones pour la resolution d’analogies entre phrases en traduction automatique par l’exemple (2020)
  • Author(s): V. Taillandier, L. Wang et Y. Lepage
  • Journal Title: In Actes de JEP-TALN-RECITAL, AFCP et ATALA, Mai 2020. HAL-ID: hal-02784759. https://www.aclweb.org/anthology/2020.jeptalnrecital-taln.9 Volume: 2
  • Peer Reviewed / Open Access / Int'l Joint Research
• [Journal Article] An approach to case-based reasoning based on local enrichment of the case base (2019)
  • Author(s): Y. Lepage and J. Lieber
  • Journal Title: Lecture notes in computer science (LNCS) Volume: 11680 Pages: 235-250
  • DOI: 10.1007/978-3-030-29249-2_16
  • ISBN: 9783030292485, 9783030292492
  • Peer Reviewed / Open Access / Int'l Joint Research
• [Journal Article] Case-Based Translation: First Steps from a Knowledge-Light Approach Based on Analogy to a Knowledge-Intensive One (2018)
  • Author(s): Y. Lepage and J. Lieber
  • Journal Title: ICCBR 2018, LNAI Volume: 11156 Pages: 273-288
  • DOI: 10.1007/978-3-030-01081-2_37
  • ISBN: 9783030010805, 9783030010812
  • Peer Reviewed / Open Access / Int'l Joint Research
• [Presentation] Semantico-formal resolution of analogies between sentences (2019)
  • Author(s): Y. Lepage
  • Organizer: Proceedings of the 9th Language & Technology Conference (LTC 2019) Human Language Technologies as a Challenge for Computer Sicience and Linguistics, pages 57-61, May 2019
  • Int'l Joint Research
• [Presentation] Iterative training for unsupervised word embedding mapping (2019)
  • Author(s): S. Yu and Y. Lepage
  • Organizer: Proceedings of the 9th Language & Technology Conference (LTC 2019) Human Language Technologies as a Challenge for Computer Science and Linguistics, pages 62-66, May 2019
  • Int'l Joint Research
• [Presentation] Numerical methods for retrieval and adaptation in Nagao’s EBMT model (2018)
  • Author(s): K. He, T. Zhao, and Y. Lepage
  • Organizer: 2018 International Conference on Advanced Computer Science and Information Systems (ICACSIS 2018)
  • Int'l Joint Research
• [Presentation] String transformations preserving analogies (2018)
  • Author(s): Y. Lepage
  • Organizer: 2018 International Conference on Advanced Computer Science and Information Systems (ICACSIS 2018)
  • Int'l Joint Research
• [Presentation] Context encoder for analogies on strings (2018)
  • Author(s): T. Zhao and Y. Lepage
  • Organizer: 32th Pacific Asia Conference on Language, Information and Computation (PACLIC 32)
• [Remarks] Kakenhi Kiban C 18K11447
  • URL: http://lepage-lab.ips.waseda.ac.jp/en/projects/kakenhi-kiban-c-18k11447/
• [Remarks] Kakenhi Kiban C 18K11447 -- Results of experiments
  • URL: http://lepage-lab.ips.waseda.ac.jp/en/projects/kakenhi-kiban-c-18k11447/