Please use this identifier to cite or link to this item:

The first Mirandese text-to-speech system 

File SizeFormat 
ferreira_et_al_2016.pdf156.34 kBAdobe PDFView/Open

Item Summary

Title: The first Mirandese text-to-speech system 
Authors: José Pedro Ferreira
Cristiano Chesi
Daan Baldewijns
Daniela Braga
Miguel Dias
show 1 moreMargarita Correia
show less
Issue Date: Feb 2016
Publisher: University of Hawai'i Press
Citation: Ferreira, José Pedro, Cristiano Chesi, Daan Baldewijns, Daniela Braga, Miguel Dias, and Margarita Correia. 2016. The first Mirandese text-to-speech system. In Vera Ferreira and Peter Bouda (eds.). Language Documentation and Conservation in Europe. 150-158. Honolulu: University of Hawai'i Press.
Series/Report no.: LD&C Special Publication
Abstract: This paper describes the creation of base NLP resources and tools for an under-resourced minority language spoken in Portugal, Mirandese, in the context of the generation of a text-to-speech system, a collaborative citizenship project between Microsoft, ILTEC, and ALM – Associaçon de la Lhéngua Mirandesa. Development efforts encompassed the compilation of a large textual corpus, definition of a complete phone-set, development of a tokenizer, inflector, TN and GTP modules, and creation of a large phonetic lexicon with syllable segmentation, stress mark-up, and POS. The TTS system will provide an open access web interface freely available to the community, along with the other resources. We took advantage of mature tools, resources, and processes already available for phylogenetically-close languages, allowing us to cut development time and resources to a great extent, a solution that can be viable for other lesser-spoken languages which enjoy a similar situation.
Sponsor: National Foreign Language Resource Center
ISBN: 978-0-9856211-5-5
Rights: Creative Commons Attribution Non-Commercial Share Alike License
Appears in Collections:LD&C Special Publication No. 9: Language Documentation and Conservation in Europe

Items in ScholarSpace are protected by copyright, with all rights reserved, unless otherwise indicated.