In principle, the longest phonological phrase possible would coincide with
the maximum time elapsed, before people would need to take a breath. However,
in practice, Phonological Phrases are never that long and speakers impose a
much finer and richer partition on the utterances they render. In the project
presented an attempt is made to explore the reasons underlying the speakers'
choice on prosodization and identify regularities and constraints primarily
diverse in nature, aiming at a more informed choice in PP prediction for Text
to Speech Synthesis.The work departs from the comparison between two different
proposals on the relation between syntactic and prosodic structure, namely:
a) the set of constraints developed by H. Truckenbrodt (1999) within an
Optimality Theory framework and building on an edge-based approach to the
conditions governing the syntactic and phonological component interface,
b) M.Steedman's (2000) theory according to which the prosody of the utterance is
determined in terms of Information Structure alone, the latter reflected on
the more flexible syntactic surface structure of Combinatory Categorical
Grammar.
All research is based on a set of approximately 600 sentences from the BU Radio
News Corpus.