A New Prosodic Phrasing Model for Indian Language Telugu

摘要

Prosodic phrasing is an important and more difficult a problem for Indian languages, as the Indian language scripts use very little or no punctuation. This paper reports a preliminary attempt on data-driven modeling of prosodic phrase boundary prediction for the Indian language Telugu. In an effort to identify meaningful features that affect the prosodic phrasing, a new feature, namely mopheme tag, is defined. A Classification and Regression Tree (CART) based data-driven phrasing model is developed for the prosodic phrase boundary prediction and the usefulness of the morpheme tag feature is further demonstrated in an evaluation process. The phrasing model developed has been implemented in an Indian language Text-to-Speech synthesis system being developed within Festival framework.