P04-1086 each other . Previous work on pitch accent prediction , how - ever , neglected the
P04-1086 been used in acoustic models of pitch accent prediction . These features include the
P00-1030 collocation-based measures for pitch accent prediction . Our initial hypothesis was
N07-1001 in the AT&T synthesizer for pitch accent prediction . Boundary tones are usually
P00-1030 machine learning helps to derive pitch accent prediction models using this feature . Finally
P04-1086 Conditional Random Fields ( CRFs ) to pitch accent prediction task in order to incorporate
N07-1001 prediction but is slightly worse for pitch accent prediction . 5.2 Supertagger performance
W03-1019 many others , such as chunking , pitch accent prediction and speech edit detection . These
P04-1086 baseline and previous models of pitch accent prediction on the Switchboard Corpus . 1
P00-1030 word collocation is useful for pitch accent prediction , we first employed Spearman
P00-1030 predictability are useful for pitch accent prediction . ` Since pointwise mutual information
N07-1002 feature in a domain independent pitch accent prediction task . Their hypothesis that
P04-1086 of rhythm and timing to model pitch accent prediction . CRFs have the theoretical advantage
W07-0206 up to 5.3 % on two real tasks : pitch accent prediction and optical character recognition
N07-1001 performs better than Festival in pitch accent prediction and the latter performs better
P04-1086 sequence corresponding to x . In pitch accent prediction , xt is a word and yt is a binary
P04-1086 Jon 's LEAVING . " ) . Accurate pitch accent prediction lies in the successful combination
P00-1030 <title> Modeling Local Context for Pitch Accent Prediction </title> Shimei Pan Julia Hirschberg
P04-1086 sequence labeling techniques to pitch accent prediction task . 8 Acknowledgements This
P04-1086 used in Hidden Markov models of pitch accent prediction have been very limited , e.g.
hide detail