Rings における二項対立の識別 : 二項対立への主要な定量的アプローチ

Division of Language and Culture, Graduate School of Humanities, The University of Osaka

二項対立

エンベディングモデル

コサイン類似度

BERTopic

binary oppositions

embeddings model

cosine distance

This paper presents a primary quantitative approach to identifying binary oppositions in the Lord of the Rings. Using a custom-built Word2Vec model trained exclusively on Tolkien’s works, and integrating it with BERTopic for topicmodelling, the study explores how semantic relationships reflect Tolkien’s narrative structure and thematic oppositions. These oppositions were revealed through analysing the results. The analysis showed binaries such as healing and trauma, and nature and war, which are examined both at a contextual and lexical level. High and low cosine similarity scores between topics are used as indicators of semantic distance, identifying oppositions both within the same thematic categories and across conceptually distinct ones. The study concludes with a discussion on the implications of an embeddings model for literary analysis, proposing further exploration across Tolkien’s legendarium to understand the broader impact of binary oppositions in his work.