Amalia Todiraºcu * A Unification-Based Model for Speech Generation



Pitch is influenced by the type of the sound pronounced. The flow of the air will be strong at the beginning of the pronunciation of an utterance and the baseline of pitch raised. The pitch falls momentarily when a sound is produced with the constriction of the vocal tract, like fricatives or stops.

Sound quantity (the length of a sound) is difficult to be expressed by using the formalism introduced before.

Sound quality can be used, because sounds can be classified in this respect, each syllable having a most prominent phoneme.

4.3. Rhythm and intonation rules

Unstressed syllables are grouped together with a stressed one in a syntagm or a sense-group. Each sense-group consists of words in close grammatical connection. The rhythmic beats of an utterance coincide with the stressed syllables and all the unaccented syllables are grouped around the stressed ones, being language-dependent. English is a stress-timed language, and Romanian is a syllable-timed language, where syllables take an almost equal amount of time. The rhythmic division is made in the following way: the first accented syllable in the utterance carries with it all the preceding unaccented syllables and the following ones up to the second accented syllable, and the process is repeated until the end of the sentence.



5. Conclusion

The model designed in this paper gives a format for phonological representations which is rigorous enough to admit computational interpretation. It uses as fundamentals a very modern formalism (HPSG) in computational linguistics, which has been applied successfully in the domain of phonology. The model tries to be general and language-independent, but the special case of Romanian language is also treated.



References

  1. C.I. POLLARD, A. SAG, Information-based Syntax and Semantics, CSLI, 1992.

  2. C.I. POLLARD, A. SAG, Head-driven Phrase Structured Grammar, CSLI, 1994.

  3. S. BIRD, E. KLEIN, Phonological Analysis in Typed Feature Systems, Computational Linguistics Journal, 20, 3 (1994).

  4. O. ANDERSEN, P. DALSGAARD, W. BARRY, Data-driven Identification of Poly- and Mono-phonemes for Four European Languages, paper presented at Eurospeech '93, Vol. 1, pp. 759-762.

  5. J.R. DE PIJPER, A. SANDERMAN, Prosodic Cues to the Perception of Constituent Boundaries, paper presented at Eurospeech'93, Vol. 2, pp. 1211-1214.

  6. W.N. CAMPBELL, C.W. WIGHTMAN, Prosodic Encoding of Syntactic Structure for Speech Synthesis, paper presented at ICSLP '94 (International Conference on Spoken Language Processing), 1992, Banffs, Alberta, Canada, Vol. 2, pp. 1167-1170.

  7. F. BEAUGENDRE, A. LACHERET-DUJOUR, Automatic Generation of French Intonation Based on a Perceptual Study and Morphosyntactic Information, paper presented at Eurospeech '93, Vol. 2, pp.1219-1222.

  8. C. GOGÃLNICEANU, The English Phonetics and Phonology, Edit. Fundaþiei "Chemarea", Iaºi, 1993.




173

Previous Index Next