Due to the interaction between lexical tones and intonation in Chinese speech melody, the contour patterns of Chinese IPs （intonational phrases） have long been an intriguing yet complicated subject. Within the AM （...Due to the interaction between lexical tones and intonation in Chinese speech melody, the contour patterns of Chinese IPs （intonational phrases） have long been an intriguing yet complicated subject. Within the AM （Autosegmental-Metrical） framework, this paper makes an empirical study on the subject based on a small-scale database of 22 standard Chinese news stories. A five-degree D value system is proposed for data normalization so that various contour patterns of IPs can be identified and compared despite complications in pitch register and pitch range. The major findings are： Chinese IPs mainly comprise 2, 3, or 4 ips （intermediate phrases）. And the contour patterns of Chinese IPs suggest that the statement intonation in Chinese is primarily indicated by the near-bottom Dmin value of the last ip. Meanwhile, when the primary indicator is not typical, the falling trend of the last part of the top-line can serve as a compensatory perceptual clue.展开更多
Automatic prosodic break detection and annotation are important for both speech understanding and natural speech synthesis. In this paper, we discuss automatic prosodic break detection and feature analysis. The contri...Automatic prosodic break detection and annotation are important for both speech understanding and natural speech synthesis. In this paper, we discuss automatic prosodic break detection and feature analysis. The contributions of the paper are two aspects. One is that we use classifier combination method to detect Mandarin and English prosodic break using acoustic, lexical and syntactic evidence. Our proposed method achieves better performance on both the Mandarin prosodic annotation corpus Annotated Speech Corpus of Chinese Discourse and the English prosodic annotation corpus -- Boston University Radio News Corpus when compared with the baseline system and other researches＇ experimental results. The other is the feature analysis for prosodic break detection. The functions of different features, such as duration, pitch, energy, and intensity, are analyzed and compared in Mandarin and English prosodic break detection. Based on the feature analysis, we also verify some linguistic conclusions.展开更多
文摘Due to the interaction between lexical tones and intonation in Chinese speech melody, the contour patterns of Chinese IPs （intonational phrases） have long been an intriguing yet complicated subject. Within the AM （Autosegmental-Metrical） framework, this paper makes an empirical study on the subject based on a small-scale database of 22 standard Chinese news stories. A five-degree D value system is proposed for data normalization so that various contour patterns of IPs can be identified and compared despite complications in pitch register and pitch range. The major findings are： Chinese IPs mainly comprise 2, 3, or 4 ips （intermediate phrases）. And the contour patterns of Chinese IPs suggest that the statement intonation in Chinese is primarily indicated by the near-bottom Dmin value of the last ip. Meanwhile, when the primary indicator is not typical, the falling trend of the last part of the top-line can serve as a compensatory perceptual clue.
文摘在语音合成系统中,语调短语的自动预测是影响合成语音的自然度和可懂度的关键因素之一。采用了最大熵(Maximum Entropy,ME)模型从无限制的文本中预测语调短语,并且提出了一个自动生成特征模板的层次聚类算法,从而减少了最大熵模型训练过程中的人工参与。实验结果表明,对于语调短语预测而言,最大熵模型明显优于分类与回归树(Classification And Regression Trees,CART)。相比手工总结的特征模板,自动生成的特征模板不仅将语调短语预测的F-score提高了3.18,而且将最大熵模型的大小缩小了78.38。
基金Supported by the National Natural Science Foundation of China under Grant Nos. 90820303,90820011the Natural Science Foundation of Shandong Province of China under Grant No. ZR2011FQ024
文摘Automatic prosodic break detection and annotation are important for both speech understanding and natural speech synthesis. In this paper, we discuss automatic prosodic break detection and feature analysis. The contributions of the paper are two aspects. One is that we use classifier combination method to detect Mandarin and English prosodic break using acoustic, lexical and syntactic evidence. Our proposed method achieves better performance on both the Mandarin prosodic annotation corpus Annotated Speech Corpus of Chinese Discourse and the English prosodic annotation corpus -- Boston University Radio News Corpus when compared with the baseline system and other researches＇ experimental results. The other is the feature analysis for prosodic break detection. The functions of different features, such as duration, pitch, energy, and intensity, are analyzed and compared in Mandarin and English prosodic break detection. Based on the feature analysis, we also verify some linguistic conclusions.
