Prosody prediction for speech synthesis using transformational rule-based learning

  title={Prosody prediction for speech synthesis using transformational rule-based learning},
  author={Cameron S. Fordyce and Mari Ostendorf},
Prediction of symbolic prosodic labels (pitch accents and phrase structure) is an important step in generating natural synthetic speech. This paper investigates a new automatically trainable procedure for combined accent and phrase prediction based on transformational rule-based learning. Experimental results on a radio news corpus show that accent prediction bene ts from phrase structure, but not vice versa, and that TRBL outperforms simple decision tree predictors. 
Highly Cited
This paper has 33 citations. REVIEW CITATIONS


Publications referenced by this paper.
Showing 1-10 of 61 references

Building a Very Large Natural Lan- guage Corpora: The Penn Treebank.

  • M. P. Marcus, B. Santorini
  • Computational Linguistics,
  • 1993
Highly Influential
5 Excerpts

Exploring the Nature of Transformational-Based Learning." In The Balancing Act: Combining Symbolic and Statistical Approaches to Language

  • L. A. Ramshaw, M. P. Marcus
  • 1996
Highly Influential
5 Excerpts

Computational Models of the Prosody/Syntax Mapping for Spoken Language Systems.

  • N. M. Veilleux
  • PhD thesis,
  • 1994
Highly Influential
6 Excerpts

Automatic Classi cation of Intonational Phrase Boundaries.

  • M. Q. Wang, J. Hirschberg
  • Computer Speech and Language,
  • 1992
Highly Influential
12 Excerpts

Modeling of Intonation for Speech Synthesis.

  • K. Ross
  • PhD thesis,
  • 1995
Highly Influential
3 Excerpts

Similar Papers

Loading similar papers…