Cross-Linguistic Studies on Lexical Differences based on Representation Learning

2022 Fiscal Year Final Research Report

• Project/Area Number: 18K11456
• 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: National Institute of Advanced Industrial Science and Technology
• Principal Investigator: Takamura Hiroya 国立研究開発法人産業技術総合研究所, 情報・人間工学領域, 研究チーム長 (80361773)
• Co-Investigator(Kenkyū-buntansha): 永田 亮 甲南大学, 知能情報学部, 准教授 (10403312) ; 川崎 義史 東京大学, 大学院総合文化研究科, 准教授 (40794756)
• Project Period (FY): 2018-04-01 – 2023-03-31
• Keywords: 語彙的変異 / 分散表現 / 深層学習 / 意味変化

Outline of Final Research Achievements

We investigated the statistical relationship between semantic difference in Roman cognates and six variables including frequency and polysemy. The degree of semantic difference was quantified using the cosine distance of the distributed representations of words. We conducted regression analysis and demonstrated that frequency is negatively correlated with semantic difference, while polysemy is positively correlated with semantic difference. We also found that morphologically complex word roots are less likely to undergo semantic change, while cognates that have been in use for a long time are more likely to undergo semantic change. We also examined how the new usage of "better off" came to be established. In addition, we investigated the lexical variation between writings by native speakers and non-native speakers.

Free Research Field

自然言語処理

Academic Significance and Societal Importance of the Research Achievements

単語分散表現を含む深層学習技術は、言語研究における新たな道具であり、それを実証する成果が得られている。これまで変化検出の研究が多かった中で、語彙的変異の要因を探った点で学術的意義が大きい。また、"better off"に関する研究では、言語学で考えられた仮説を検証しており、自然言語処理技術の言語学への貢献の形として、良い例となるだろう。