[1K1-GS-6] Language media processing: evaluation / anaysis

[1K1-GS-6-05] Diachronic Connotation Analysis of English and American Literature Based on the Similarity Between Word Vectors

Chronological Change of Connotation based on Quantification in the Works of Herman Melville

〇Jun Kaneko¹, Takashi Otsuki², Takayuki Sakaguchi³ (1. Mie University, Faculty of Education, 2. Yamagata University, Graduate School of Science and Engineering, 3. Yamagata University, Faculty of Education, Art and Science)

[[Online]]

Keywords:Natural Language Processing, English and American Literature, Word2vec, fastText, spaCy

The previous study revealed that the differences in the meanings of words could be viewed not qualitatively but rather, quantitatively. Based on this fact, the current study will attempt to verify that the nuances of words have changed over time. To clarify this, an experiment was conducted using the text of Herman Melville’s 4 novels, which ranged from books written early in his life to those written in his later years. By comparing word vectors of general linguistic data (implementing the pretrained model of fastText) with that of texts written by Melville (also using fastText), the writer’s changing nuance of words was revealed. The words selected for analysis are some important keywords based on what has been already studied so far in terms of American Literature. The experiment showed that the words have changed over time in their nuances.