Title | ||
---|---|---|
Sliding Window Smoothing For Maximum Entropy Based Intonational Phrase Prediction In Chinese |
Abstract | ||
---|---|---|
In Chinese TTS (Text-To-Speech) system, Intonational phrase prediction has great influence on naturalness of synthesized speech. Different kinds of statistic models have been applied to this domain, and achieved good performance. In this paper, we first build a maximum entropy model to yield the probability of each word boundary to be an intonational phrase break, and then a sliding window smoothing algorithm is proposed, in which the length distribution curve of intonational phrase acts as the sliding window. The maximum entropy model and the distribution curve are trained from 19,000 sentences and tested on a test set of 1,000 sentences. Experiment results shows that, the sliding window smoothing algorithm makes an improvement of 5.3% in terms of F-Score, 10.0% in terms of average score, and 55.6% in terms of unacceptable rate. From the results, we draw the conclusion that the length distribution information is of great usefulness for intonational phrase break prediction, and the sliding window smoothing method is quite effective to improve the performance significantly. © 2005 IEEE. |
Year | DOI | Venue |
---|---|---|
2005 | 10.1109/ICASSP.2005.1415106 | ICASSP (1) |
Keywords | Field | DocType |
statistical distributions,intonational phrase prediction,statistical models,natural languages,smoothing methods,speech synthesis,maximum entropy methods,prediction theory,sliding window smoothing,intonational phrase break prediction,chinese text-to-speech systems,maximum entropy,phrase length distribution,entropy,text to speech,maximum entropy model,probability,statistical model,testing,artificial neural networks,sliding window,statistics | Speech synthesis,Sliding window protocol,Pattern recognition,Computer science,Phrase,Speech recognition,Probability distribution,Smoothing,Artificial intelligence,Statistical model,Principle of maximum entropy,Test set | Conference |
Volume | Issue | ISSN |
1 | null | 1520-6149 |
ISBN | Citations | PageRank |
0-7803-8874-7 | 3 | 0.46 |
References | Authors | |
5 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Jian-Feng Li | 1 | 3 | 0.46 |
Guoping Hu | 2 | 309 | 37.32 |
Ren-Hua Wang | 3 | 344 | 41.36 |
Li-Rong Dai | 4 | 1070 | 117.92 |