書誌：電子情報通信学会技術研究報告 . WIT, 福祉情報工学,2011
電子情報通信学会技術研究報告 . WIT, 福祉情報工学 111(226), 67-72, 2011-09-29
大川 万里子, 大橋 浩輝, 市川 熹
Abstract: Overlapping speech frequently occurs in the dialogues between native speakers. According to previous studies on Japanese or English dialogues, the frequency of overlapping speech is over 40 percent in the Maptask corpus. We assume that the frequency of overlaps is related to the smooth conversation. We analyzed the two types of Japanese dialogues: between a non-native (Chinese) and a native speaker, and between non-native speakers. We except that non-native speakers' utterances would overlap less often than native speakers' ones because a preceding speaker has difficulty predicting when does a following speaker begin utterances in non-native language. As a result, contrary to our expectation, the frequency of overlaps in the Japanese dialogues between a non-native speaker and a native speaker is larger than that between native speakers.