作者: Sadao Kurohashi , Masashi Okamoto , Nobuhiro Kaji
DOI:
关键词:
摘要: There are a lot of differences between expressions used in written language and spoken language. It is one the reasons why speech synthesis applications prone to produce unnatural speech. This paper represents method paraphrasing unsuitable for into suitable ones. Those two can be distinguished based on occurrence probability corpora which automatically collected from Web. Experimental results indicated effectiveness our method. The precision was 94%, accuracy learning paraphrases 76 %.