EI / SCOPUS / CSCD 收录

中文核心期刊

NI Chongjia, LIU Wenju, XU Bo. The comparison and analysis between Chinese Putonghua and English stress automatic annotation[J]. ACTA ACUSTICA, 2012, 37(5): 553-560. DOI: 10.15949/j.cnki.0371-0025.2012.05.012
Citation: NI Chongjia, LIU Wenju, XU Bo. The comparison and analysis between Chinese Putonghua and English stress automatic annotation[J]. ACTA ACUSTICA, 2012, 37(5): 553-560. DOI: 10.15949/j.cnki.0371-0025.2012.05.012

The comparison and analysis between Chinese Putonghua and English stress automatic annotation

  • Although Chinese Putonghua stress automatic annotation and English pitch accent automatic annotation have been studied extensively, there is seldom works on exploring the comparison between Putonghua stress automatic annotation and English pitch accent automatic annotation. In this paper, we make comparison between Putonghua stress automatic annotation and English pitch accent automatic annotation based on Putonghua prosodic annotation corpus--ASCCD and English prosodic annotation corpus--Boston University Radio News Corpus, and examine the generalization of different features in different corpora. Through the experiments of stress automatic annotation based on boosting classification and regression tree, feature analysis and the acoustic comparison of stress automatic annotation based on the mutual information~ we get the following conclusions: (1) The correct rate in Putonghua stress automatic annotation is lower than that of English pitch accent automatic annotation; (2) When annotating stress, the lexical and syntactic related features are more important than the acoustic related features; (3) The different acoustic features have different function on stress annotation, and in all these features, the duration related features are important to Putonghua and English stress automatic annotation; (4) The mutual information value of most features in English is higher than that of in Putonghua.
  • loading

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return