Automatic Checking of Foreign Language Writing Based on Neural Machine Translation and Natural Language Inference

2022 Fiscal Year Final Research Report

• Project/Area Number: 20K00830
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 02100:Foreign language education-related
• Research Institution: Shizuoka University
• Principal Investigator: Tsunakawa Takashi 静岡大学, 情報学部, 講師 (30611214)
• Project Period (FY): 2020-04-01 – 2023-03-31
• Keywords: 添削 / 和文英訳 / 外国語作文 / 誤り検出 / 誤り訂正

Outline of Final Research Achievements

This study aims to construct a translation correction system that enables users to effectively self-study foreign language composition based on recent natural language processing technologies based on deep learning. In particular, we focused on the detection and correction of semantic errors in the sentences. We constructed an automatic correction system with natural language inference, sentence similarity estimation, and word alignment based on the state-of-the-art natural language processing technologies. In the experiment, the proposed system achieved over 70% in both of precision and recall for detecting inappropriate sentences, and about half of the correction results were evaluated by humans as being valid.

Free Research Field

自然言語処理

Academic Significance and Societal Importance of the Research Achievements

和文英訳問題は英語学習者にとって重要な部分を占めている一方で、学習者自身で回答の正誤を判定することはしばしば困難である。文法的誤りの検出・訂正においては実用化されたシステムが存在する一方で、意味的誤りの検出・訂正はまだ研究途上にあった。本研究は最新の自然言語処理技術を適用することで、任意の和文英訳問題に対する意味的誤りの検出と添削がある程度の性能で可能であることを示した上で、現状残っている学術的課題を明らかにした。