Predicting Prosodic Phrasing Using Linguistic Features

The prosodic structure of speech is based on complex interaction within and between several different levels of linguistic, and paralinguistic organization, and is expressed in the modulation of F0, intensity, duration, and voice quality, as well as the occurrence of pauses. Even though leading theories of prosody maintain that prosody is shaped through the interaction of grammatical factors from phonology, syntax, semantics, and pragmatics [1][2][3][4], there is no consensus on how to model their interaction. I provide a new probabilistic model of the mapping between prosody and phonology, syntax, and argument structure. The model encodes phonological features, shallow syntactic constituent structure, and basic argument structure. A machine learning experiment using these features to predict prosodic phrase boundaries achieves more than 92% accuracy in predicting prosodic boundary location: 86.10% precision and recall in predicting boundary locations and 94.61% in predicting locations where no boundary is present. An experiment for predicting the strength of prosodic boundaries achieve 88.06% accuracy. This study sheds light on the relationship between prosodic phrase structure and other grammatical structures.