Emotional Science of Literatures

2023 Fiscal Year Final Research Report

• Project Area: Restructuring human sciences based on decoding of emotional information
• Project/Area Number: 21H05061
• Research Category: Grant-in-Aid for Transformative Research Areas (B)
• Allocation Type: Single-year Grants
• Review Section: Transformative Research Areas, Section (I)
• Research Institution: The Institute of Statistical Mathematics
• Principal Investigator: Mochihashi Daichi 統計数理研究所, 統計基盤数理研究系, 准教授 (80418508)
• Co-Investigator(Kenkyū-buntansha): 大関 洋平 東京大学, 大学院総合文化研究科, 講師 (10821994) ; 内海 彰 電気通信大学, 大学院情報理工学研究科, 教授 (30251664) ; 小林 一郎 お茶の水女子大学, 基幹研究院, 教授 (60281440)
• Project Period (FY): 2021-08-23 – 2024-03-31
• Keywords: 自然言語処理 / 脳科学 / 感情情報処理

Outline of Final Research Achievements

In this project, we investigated how people feel when they are exposed to language art such as tanka through both neuroscientific measurements and statistical natural language processing. fMRI analysis of brain activity revealed that the occipital lobe and Broca's area in particular contribute to the judgment of whether a text is poetic or not, and that linguistic characteristics also differ quantitatively between plain sentences and poetic sentences. The results also showed that there were quantitative differences between plain and poetic sentences. By statistical analysis using psychometrics, we proposed a statistical model that can measure the potential characteristics and goodness of tanka based only on the evaluation data of tanka, and can quantify the different evaluation axes among the evaluators in this model. It was also found that left-corner progressive parsing is a good predictor of brain activity when humans read sentences.

Free Research Field

自然言語処理、機械学習

Academic Significance and Societal Importance of the Research Achievements

本研究課題により、短歌のような言語芸術の鑑賞およびその評価過程を、脳科学および統計的分析を通じて客観化・定量化できることが明らかになった。言語芸術に関する自然言語処理は、特に日本においては研究が少なく、今後の研究に重要なマイルストーンを提供したと考えている。