• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to previous page

Machine Translation based on Multimodal Quality Estimation

Research Project

Project/Area Number 19K21533
Project/Area Number (Other) 18H06465 (2018)
Research Category

Grant-in-Aid for Research Activity Start-up

Allocation TypeMulti-year Fund (2019)
Single-year Grants (2018)
Review Section 1001:Information science, computer engineering, and related fields
Research InstitutionOsaka University

Principal Investigator

Kajiwara Tomoyuki  大阪大学, データビリティフロンティア機構, 特任助教(常勤) (70824960)

Project Period (FY) 2018-08-24 – 2020-03-31
Project Status Completed (Fiscal Year 2019)
Budget Amount *help
¥2,730,000 (Direct Cost: ¥2,100,000、Indirect Cost: ¥630,000)
Fiscal Year 2019: ¥1,300,000 (Direct Cost: ¥1,000,000、Indirect Cost: ¥300,000)
Fiscal Year 2018: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
Keywords機械翻訳 / マルチモーダル / 知能情報学 / 自然言語処理 / 品質推定
Outline of Research at the Start

2020年の東京五輪に向けて、機械翻訳の精度向上が急務である。従来の機械翻訳は、翻訳文と正解文の単語単位の一致率を最大化する訓練を行ってきたが、この正解文は数ある正しい翻訳の表現の一例に過ぎない。本研究では、マルチモーダル品質推定によって正解文の表現に依存しない訓練を行い、機械翻訳を高度化する。本研究の背景には、対訳などの大規模なテキストとテキストの対応データが一部の言語でしか利用できない一方で、SNSの普及により画像とテキストの対応データが多くの言語で大規模に利用できるという利点がある。

Outline of Final Research Achievements

Improving machine translation performance is an urgent task for the Tokyo Olympics. In this research, we worked on automatic evaluation of machine translation and multimodal machine translation.
In machine translation training, translated sentences that differ superficially from reference sentences are penalized even if they are semantically correct. Therefore, we proposed an automatic evaluation method using vector representation of sentences, and achieved high correlation with human evaluation while reducing the dependence on reference sentences.
In conventional multimodal machine translation, images are divided into uniform sizes, which makes matching with text difficult. Therefore, we proposed a multimodal machine translation method using images divided into semantic units and improved the translation quality.

Academic Significance and Societal Importance of the Research Achievements

東京五輪や大阪万博に向けて、機械翻訳の精度向上が急務である。機械翻訳モデルの効率的な改善のために、人手評価との高い相関を持つ自動評価手法の開発が重要である。本研究では、単語やフレーズの一致といった局所的な情報に頼っていた自動評価を改善し、文全体の大域的な情報をもとに人手評価との高い相関を持つ自動評価手法を構築した。機械翻訳に関する国際会議WMT-2018において開催された自動評価手法の性能を競うコンペティションにおいては、ドイツ語から英語や中国語から英語などの全7言語対において、我々の提案手法が世界最高性能を達成した。

Report

(3 results)
  • 2019 Annual Research Report   Final Research Report ( PDF )
  • 2018 Annual Research Report
  • Research Products

    (4 results)

All 2020 2019 2018

All Journal Article (1 results) (of which Peer Reviewed: 1 results,  Open Access: 1 results) Presentation (3 results) (of which Int'l Joint Research: 2 results)

  • [Journal Article] Metric for Automatic Machine Translation Evaluation based on Pre-trained Sentence Embeddings2019

    • Author(s)
      嶋中宏希, 梶原智之, 小町守
    • Journal Title

      Journal of Natural Language Processing

      Volume: 26 Issue: 3 Pages: 613-634

    • DOI

      10.5715/jnlp.26.613

    • NAID

      130007761392

    • ISSN
      1340-7619, 2185-8314
    • Year and Date
      2019-09-15
    • Related Report
      2019 Annual Research Report
    • Peer Reviewed / Open Access
  • [Presentation] Double Attention-based Multimodal Neural Machine Translation with Semantic Image Regions2020

    • Author(s)
      Yuting Zhao, Mamoru Komachi, Tomoyuki Kajiwara, Chenhui Chu
    • Organizer
      Proceedings of the 22nd Annual Conference of the European Association for Machine Translation (EAMT 2020)
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] BERTを用いた機械翻訳の自動評価.2019

    • Author(s)
      嶋中宏希, 梶原智之, 小町守.
    • Organizer
      言語処理学会第25回年次大会
    • Related Report
      2018 Annual Research Report
  • [Presentation] RUSE: Regressor Using Sentence Embeddings for Automatic Machine Translation Evaluation.2018

    • Author(s)
      Hiroki Shimanaka, Tomoyuki Kajiwara, Mamoru Komachi.
    • Organizer
      Proceedings of the Third Conference on Machine Translation (WMT 18)
    • Related Report
      2018 Annual Research Report
    • Int'l Joint Research

URL: 

Published: 2018-08-27   Modified: 2024-03-26  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi