Automatic Prosodic Labeling with Conditional Random Fields and Rich Acoustic Features

Many acoustic approaches to prosodic labeling in English have employed only local classifiers, although text-based classification has employed some sequential models. In this paper we employ linear chain and factorial conditional random fields (CRFs) in conjunction with rich, contextually-based prosodic features, to exploit sequential dependencies and to facilitate integration with lexical features. Integration of lexical and prosodic features improves pitch accent prediction over either feature set alone, and for lower accuracy feature sets, factorial CRF models can improve over linear chain based prediction of pitch accent.