Prosodic modeling for Mandarin Chinese speech embedded with English spelling

Abstract

Since the letter-by-letter spelling of English words or acronyms appears in Mandarin Chinese speech quite often, a prosodic modeling approach for Mandarin speech mixed with English spelling is proposed to help understand the influence of embedded foreign language and make the foreign language congruous with the primary language's prosody. Two important prosodic features, duration and pitch-mean (as a role of intonation), are discussed based on an English-Mandarin bilingual speech database, in which all the English words are in spelling style, that is, are read letter by letter. This approach considers several additive affecting factors that contribute to the variations of duration and pitch-mean. The parameters of the two modeling units were automatically estimated using the expectation-maximization (EM) algorithm. Experimental results showed that the root mean squared errors (RMSEs) obtained in the training and test sets were 8.93 and 9.00 ms in the reconstructed duration, and 0.41 and 0.83 ms in the pitch-mean respectively. This model provides a way to separate the effects of several major factors. All of the inferred weights values of the affecting factors were in close agreement with our prior linguistic knowledge. In addition, the model can provide useful cues to determine the prosodic phrase boundaries, including those occurring at intersyllable locations, with or without punctuation marks.