Maximum Entropy motivated grapheme-to-phoneme, stress and syllable boundary prediction for Portuguese text-to-speech. Barros, M. J. and Weiss, C. In Buera, L.; Lleida, E.; Miguel, A.; and Ortega, A., editors, IV Jornadas en Tecnología del Habla, pages 177-182, Zaragoza. Universidad de Zaragoza - Red Temática en Tecnologías del Habla.
abstract   bibtex   
In this paper we present a framework for grapheme- to-phoneme (G2P) conversion, stress and syllable boundary prediction for European Portuguese (EP) Text-to-Speech (TTS) Systems. For all prediction tasks Maximum-Entropy models were used for classification. Due to the need of expensive work by experts to implement rule based G2P converters there was interest in developing probabilistic models with the Maximum- Entropy approach to solve the previous mentioned symbolic pre-processing within a TTS system. The system presented in this work is a fast and flexible approach which gives good results in each of the prediction tasks, optimal for fast application development in the TTS domain. The data used for training the G2P conversion model is manually labelled from continuous speech with natural vocalic reduction and co-articulation between words effects, common in Portuguese continuous speech. The framework is used for EP but is also usable for Brazilian Portuguese (BP) where minor changes have to be done in the G2P training data whereas stress and syllable models are the same.
@inproceedings{barros_maximum_2006,
	Address = {Zaragoza},
	Author = {Barros, Maria João and Weiss, Christian},
	Booktitle = {IV Jornadas en Tecnología del Habla},
	Date = {2006},
	Date-Modified = {2016-09-25 21:28:15 +0000},
	Editor = {Buera, Luis and Lleida, Eduardo and Miguel, Antonio and Ortega, Alfonso},
	File = {Attachment:files/884/Barros, Weiss - 2006 - Maximum Entropy motivated grapheme-to-phoneme, stress and syllable boundary prediction for Portuguese text-to-spe.pdf:application/pdf},
	Keywords = {automatic phonetic transcription, lexical stress, phonetics, Portuguese, prosody, speech synthesis, speech technology, text-to-speech, transcription},
	Pages = {177-182},
	Publisher = {Universidad de Zaragoza - Red Temática en Tecnologías del Habla},
	Title = {Maximum Entropy motivated grapheme-to-phoneme, stress and syllable boundary prediction for Portuguese text-to-speech},
	Abstract = {In this paper we present a framework for grapheme- to-phoneme (G2P) conversion, stress and syllable boundary prediction for European Portuguese (EP) Text-to-Speech (TTS) Systems. For all prediction tasks Maximum-Entropy models were used for classification. Due to the need of expensive work by experts to implement rule based G2P converters there was interest in developing probabilistic models with the Maximum- Entropy approach to solve the previous mentioned symbolic pre-processing within a TTS system. The system presented in this work is a fast and flexible approach which gives good results in each of the prediction tasks, optimal for fast application development in the TTS domain. The data used for training the G2P conversion model is manually labelled from continuous speech with natural vocalic reduction and co-articulation between words effects, common in Portuguese continuous speech. The framework is used for EP but is also usable for Brazilian Portuguese (BP) where minor changes have to be done in the G2P training data whereas stress and syllable models are the same.},
	Bdsk-File-1 = {YnBsaXN0MDDUAQIDBAUGJCVYJHZlcnNpb25YJG9iamVjdHNZJGFyY2hpdmVyVCR0b3ASAAGGoKgHCBMUFRYaIVUkbnVsbNMJCgsMDxJXTlMua2V5c1pOUy5vYmplY3RzViRjbGFzc6INDoACgAOiEBGABIAFgAdccmVsYXRpdmVQYXRoWWFsaWFzRGF0YV8QXS4uLy4uLy4uL0JpYmxpb2dyYWZpYS9QYXBlcnMvQmFycm9zL01heGltdW0gRW50cm9weSBtb3RpdmF0ZWQgZ3JhcGhlbWUtdG8tcGhvbmVtZSwgc3RyZXNzLnBkZtIXCxgZV05TLmRhdGFPEQJOAAAAAAJOAAIAAAxNYWNpbnRvc2ggSEQAAAAAAAAAAAAAAAAAAADL9h/OSCsAABCGZxofTWF4aW11bSBFbnRyb3B5IG1vIzEwODY2NzFDLnBkZgAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAEIZnHNQJ0vgAAAAAAAAAAAADAAQAAAkgAAAAAAAAAAAAAAAAAAAABkJhcnJvcwAQAAgAAMv2A64AAAARAAgAANQJttgAAAABABQQhmcaEIZljgAF/EcABfuYAADARgACAGRNYWNpbnRvc2ggSEQ6VXNlcnM6AGpvYXF1aW1fbGxpc3RlcnJpOgBCaWJsaW9ncmFmaWE6AFBhcGVyczoAQmFycm9zOgBNYXhpbXVtIEVudHJvcHkgbW8jMTA4NjY3MUMucGRmAA4AdAA5AE0AYQB4AGkAbQB1AG0AIABFAG4AdAByAG8AcAB5ACAAbQBvAHQAaQB2AGEAdABlAGQAIABnAHIAYQBwAGgAZQBtAGUALQB0AG8ALQBwAGgAbwBuAGUAbQBlACwAIABzAHQAcgBlAHMAcwAuAHAAZABmAA8AGgAMAE0AYQBjAGkAbgB0AG8AcwBoACAASABEABIAbFVzZXJzL2pvYXF1aW1fbGxpc3RlcnJpL0JpYmxpb2dyYWZpYS9QYXBlcnMvQmFycm9zL01heGltdW0gRW50cm9weSBtb3RpdmF0ZWQgZ3JhcGhlbWUtdG8tcGhvbmVtZSwgc3RyZXNzLnBkZgATAAEvAAAVAAIAGP//AACABtIbHB0eWiRjbGFzc25hbWVYJGNsYXNzZXNdTlNNdXRhYmxlRGF0YaMdHyBWTlNEYXRhWE5TT2JqZWN00hscIiNcTlNEaWN0aW9uYXJ5oiIgXxAPTlNLZXllZEFyY2hpdmVy0SYnVHJvb3SAAQAIABEAGgAjAC0AMgA3AEAARgBNAFUAYABnAGoAbABuAHEAcwB1AHcAhACOAO4A8wD7A00DTwNUA18DaAN2A3oDgQOKA48DnAOfA7EDtAO5AAAAAAAAAgEAAAAAAAAAKAAAAAAAAAAAAAAAAAAAA7s=},
	Bdsk-Url-1 = {http://lorien.die.upm.es/~lapiz/rtth/JORNADAS/IV/4jth_cdrom.html}}
Downloads: 0