Differences in Speaker Individualising Information between Case Particles and Fillers in Spoken Japanese

Differences in Speaker Individualising Information between Case Particles and Fillers in Spoken Japanese Shunichi Ishihara Department of Linguistics Australian National University shunichi.ishihara@anu.edu.au Abstract This study investigates idiosyncrasy manifest- ed in language use in spoken Japanese. For this purpose, we use speaker classification techniques as analytical tools. More precisely, focusing on Japanese case particles and fillers, of which the linguistic functions differ signifi- cantly, we aim to investigate 1) the extent of speaker idiosyncrasy in the selection of certain case particles/fillers over others in Japanese monologues, and 2) the differences, if any, be- tween case particles and fillers in the degree of speaker-individualising information. We dis- cuss what contributes to the identified differ- ences between case particles and fillers. This study will contribute to the further develop- ment of automatic speaker recognition systems and authorship analysis studies. 1 Introduction We intuitively know that different people speak/write differently, even when they try to convey the same message. We also know that people tend to use their individually-selected, preferred words despite the fact that, in principle, they can use any word at any time from the vo- cabulary built up over the course of their lives. Every speaker of a given language has their own distinctive and individual version of the language – which is often referred to as their idiolect (Halliday et al. 1964). Linguistic idiosyncrasy has been studied in both spoken and written languages (yet, more extensively on written languages) (Baayen et al. 1996, Burrows 1987, Doddington 2001, Ishihara and Kinoshita 2010, Weber et al. 2002). Many of these studies were based on the unique lexical usage of authors (Holmes et al. 2001, Juola and Baayen 2005), assuming that word selection is unique to the individual author, and that their preferred selection is consistent over time (Holmes 1992). In particular, function words are often used as a feature to quantify the unique lexical usage of individual authors, and it has been attested that function words carry author- individualising information (Binongo 2003, Holmes, et al. 2001). In addition to function words, fillers (such as English “um”, “you know”, and “like”), which are unique to spoken languages, have also been reported to carry speaker idiosyncratic information (Ishihara and Kinoshita 2010, Weber, et al. 2002). Although the above studies demonstrated that function words and fillers carry speaker/writer idiosyncratic information, the de- gree/characteristics of the individualising infor- mation that they carry may be different as the type of linguistic information they provide is significantly different. We will investigate this in this study. For that purpose, we use case- particles and fillers appearing in Japanese mono- logues. Case particles are representative function words in Japanese. We use Japanese monologues because many of the previous studies used Eng- lish as the target language, and research on idio- syncrasy in spoken languages are relatively few- er than those on written languages. That being said, the current study will investi- gate 1) the extent of speaker idiosyncrasy in the selection of certain case particles/fillers over oth- ers in Japanese …

Content truncated to 500 words