AI-ASSISTED TEXT MINING OF GLOBAL ANIME FANDOM:

Abstract

Anime theme songs (anisongs) serve as a major point of contact through which non-Japanese speakers engage with the Japanese language and culture. A previous qualitative study analyzing approximately 800 YouTube comments on five globally prominent anisongs proposed a "Multilayer Engagement Model," suggesting that non-Japanese speakers' engagement unfolds across four layers: acoustic, affective, symbolic, and learning. However, as this is an exploratory study based on a limited dataset, large-scale empirical validation is required to establish the model's validity and generalizability.

Drawing on Iwabuchi's (2002) concept of cultural odorlessness as a theoretical framework, this study examines a large-scale corpus of YouTube comments on multiple anisongs spanning the 1990s to the 2020s. Employing a mixed-methods approach that combines AI-assisted text mining with qualitative analysis, the study aims to verify and refine the Multilayer Engagement Model.

The quantitative strand involves AI-assisted text classification, frequency analysis, and correlation analysis with song attributes. The qualitative strand involves close reading of representative comments to explore the relationship between cultural engagement and Japanese language engagement. A cross-temporal comparison further examines how AI-mediated digital media environments and evolving fandom cultures have shaped patterns of engagement across decades.

This study seeks to empirically clarify the mechanisms through which anisongs function as cultural catalysts independent of translation in the age of AI, offering new perspectives for research on cultural globalization and language learning motivation.