Paper short abstract:

Through investigating spontaneous and read speech, we show that the disfluent speech by Japanese native speakers has regularities, and that studying it would enable Japanese language learners to replace the disfluent quality of their own speech with a natural disfluency like that of native speakers.

Paper long abstract:

In Europe, which is geographically distant from Japan, Japanese learners don't have many opportunities to be exposed to spoken Japanese. In order to teach them spoken Japanese efficiently, we firstly need to know its regularities. In this presentation we shed light on regularities in Japanese native speakers' speech from viewpoint of disfluency by using two kinds of corpus.

The first kind is "My Funny Talk" corpus, a corpus of more or less spontaneous Japanese speech. It has been constructed by the authors since 2010, and now it includes about 200 native speakers' speech and 130 non-native speakers' speech. Another kind is a corpus of 25 Japanese-learners' speech we collected at Ankara University in 2016. It consists of read speech of a transcription of a native speaker's disfluent speech. By using these two corpora, we argue the following three points.

Firstly, the speech of native speakers basically is disfluent, with fluent speech like that of a professional announcer being the exception rather than the rule. This agrees with the understanding of recent research on description language.

Secondly, while the speech of learners also is disfluent, it differs essentially from the disfluency of the speech of native speakers. The disfluent speech by native speakers has its own regularities, while that of learners does not.

And thirdly, except when particularly excited native speakers tend to speak piecemeal in units of syntagma (bunsetsu in Japanese), and when doing so, a particular jumping intonation (e.g., "Sore de desu nee": LHHLLHL) often is observed. In contrast, many learners seem to be aiming for the goal of speaking fluently like a newscaster—something difficult even for native speakers to do—and as a result the flow of their speech tends to be interrupted here and there. It is likely that studying the systematic nature of disfluent speech by native speakers would enable Japanese language learners to replacing the disfluent quality of their own speech with a natural disfluency like that of native speakers rather than eliminating it.