Study of automatic style transformation of spoken transcripts based on statistical modeling of spontaneous speech

点击翻译按钮获取中文摘要

To enhance the availability of spontaneous speech recognition, we conducted a study of automatic transformation of spoken transcripts to the document style. Since spoken transcripts contain a variety of colloquial and redundant expressions, and no punctuation marks are inserted, readability of such transcripts is quite low. In this study, we specifically focused on transformation of text styles by modeling the correspondence between spoken and documented texts, and automatic punctuation for spoken transcripts.