Modelling Pronunciation in Discourse Context
2006 (English)In: Lund University, Centre for Languages and Literature, General Linguistics, Phonetics, Working Papers, 52. 2006, Proceedings from Fonetik 2006, Lund, June 7–9, 2006 / [ed] Gilbert Ambrazaitis and Susanne Schötz, Lund, Sweden, 2006, 69-72 p.Conference paper (Other academic)
This paper describes a method for modelling phone-level pronunciation in discourse context.Spoken language is annotated with linguistic and related information in several layers. Theannotation serves as a description of the discourse context and is used as training data fordecision tree model induction. In a cross validation experiment, the decision tree pronunciationmodels are shown to produce a phone error rate of 8.1% when trained on all availabledata. This is an improvement by 60.2% compared to using a phoneme string compiled fromlexicon transcriptions for estimating phone-level pronunciation and an improvement by42.6% compared to using decision tree models trained on phoneme layer attributes only.
Place, publisher, year, edition, pages
Lund, Sweden, 2006. 69-72 p.
Computer Science Language Technology (Computational Linguistics)
IdentifiersURN: urn:nbn:se:kth:diva-51894OAI: oai:DiVA.org:kth-51894DiVA: diva2:465188
Fonetik 2006, Lund, Sweden, June 7–9, 2006
tmh_import_11_12_14. QC 201201182011-12-142011-12-142012-01-18Bibliographically approved