Quantitative Analysis of Literary Works Novels of Sir Arthur Conan Doyle

ショウセツ　テクスト　ノ　ケイリョウテキ　ブンセキ　アーサー　コナン　ドイル　ノ　サクヒン　カラ

クロダ, アヤカ

Kuroda, Ayaka

Graduate School of Language and Culture, Osaka University

オオサカダイガク　ダイガクイン　ゲンゴブンカ　ケンキュウカ

This study attempts to provide a new perspective for literary studies through quantitative investigation of words in texts with special reference to word frequency patterns. Two types of machine-learnin g analyses are conducted to find differences between historical fiction and detective fiction of Sir Arthur Conan Doyle. While Conan Doyle is well-known for the Sherlock Holmes series, his strong inclination for historical fiction has hardly been recognized. A number of studies have carried out to examin e personalities of characters or estimate the dates of composition for some of the texts that belong to the Holmes series. Few studies, however, have focused on Doyle’s hi sto rical fiction. Still less critical attention has been paid to stylistic aspects of his novels and short stories. Machine-learning approaches made it possible to highlight linguistic/stylistic features that distinguish Doyle’s historical fiction from his detective fiction. We used Random forests to show genre-specific‘keywords’， or words with a high keyness value so as to discriminate between the two categories of texts. MALLET was used in conjunction to build topic models based on Latent Dirichle t allocation (LDA). What emerges from our analyses are linguistic features that differentiate between the two text genres.

Quantitative Analysis of Literary Works Novels of Sir Arthur Conan Doyle

ショウセツ　テクスト　ノ　ケイリョウテキ　ブンセキ　アーサー　コナン　ドイル　ノ　サクヒン　カラ

クロダ, アヤカ

Kuroda, Ayaka

Graduate School of Language and Culture, Osaka University

オオサカダイガク　ダイガクイン　ゲンゴブンカ　ケンキュウカ