Cross-linguistic variation in turn taking practices: A computational study of the Callhome corpus.

摘要

This study uses transcripts of the Callhome corpus of phone conversations in Arabic, English, German, Mandarin and Japanese to do a cross-linguistic comparison of various aspects of turn-taking practices. Specifically it is looking at the lengths of utterances and the frequency of "continuers" (per Schegloff, 1982) across the languages of the corpus, as well as the frequency of speaker overlap. The study uses a computational method, which allows for a survey of a very large amount of data across many languages. Findings include the fact that Japanese makes the most frequent use of continuers out of the languages studied, and Arabic the least. The study also found that the trends in utterance length do vary significantly between all the languages of the corpus, and that the variation is partially due to factors other than the frequency of continuers. The language with the longest utterances was Spanish, and the shortest was Mandarin, followed closely by Japanese. The frequency of overlapping speech in all languages was high, although English was significantly lower than many other languages.