An Exploratory Investigation of IPA Texts with LDA: Using Alfred Tennyson's Poems

LDA　トピック　モデル　ニヨル　IPA　テキスト　ブンセキ　ノ　ココロミ　アルフレッド・テニスン　ノ　インブン　ヲ　モチイテ

フジタ, イク

Fujita, Iku

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

LDA

国際音声字母

アルフレッド・テニスン

詩

topic model

latent Dirichlet allocation

International Phonetic Alphabets

Alfred Tennyson

Poetry

This paper aims to investigate whether Latent Dirichlet Allocation(LDA) topic modelling can also detect the rhymes in poetry. LDA is a stylometric method that detects hidden semantic structures in texts and is now used primarily to examine prose texts. Among a relatively small number of studies on poems using LDA, Navarro-Colorado(2018) suggested that LDA can also detect rhymes in Spanish-written poetry corpus. However, speaking of the Einglish language, the spellings and the pronounciations/sounds do not always work correspond. Additionally, numerous rhymes can be found structured by the units of syllables or phonemes. Thus, applying LDA on raw English texts may not be the optimised method for detectig rhymes in texts. Considering the dissidence between spelling and sound of Einglish, this study analyses two texts-types, written in International Phonetic Alphabets(IPA) and a text of which all words in IPA are devided into vowels and consonants. Through the analysis and discussion of the emerging resurlts of LDA, this paper provides the possibility of employing LDA on IPA texts as well as issue of analysing IPA and more small-unit texts.