Ambiente web de suporte à transcrição fonética automática de lemas em verbetes de dicionários do português do Brasil

• Ano de defesa: 2015
• Autor(a) principal: Serrani, Vanessa Marquiafável [UNESP]
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Linguistica aplicada; Língua portuguesa - Brasil - Lexicografia; Lingua portuguesa - Fonetica; Linguagens e línguas Transcrições fonéticas; Linguistica - Processamento de dados; Applied linguistics
• Link de acesso: http://hdl.handle.net/11449/138398; http://www.athena.biblioteca.unesp.br/exlibris/bd/cathedra/02-05-2016/000864228.pdf

Automatic Phonetic Transcription is a crucial task for many applications of different areas. Besides Text-to-Speech systems, which need high quality transcriptions in order to have a good performance, automatic phonetic transcription tools can be useful in many areas (phonetics, phonology, language learning, and so on) in order to obtain preliminary transcriptions of large corpora. This kind of system often generate a phonetic transcription directly from the orthographic form. Considering the effort expended by lexicographers in the development of manual phonetic transcriptions to be added in their dictionaries, this study aimed the creation of a computational environment to support automatic phonetic transcription of lemmas, from mono or multilingual Brazilian dictionaries entries, to be available free of charge, via the Web, called PETRUS - PhonEtic TRanscriber for User Support. We intend to contribute, even modestly, so that lexicographers, our main target audience, can enjoy the benefits of automating the phonetic transcription, reducing the time and difficulty in integrating this kind of information in their lexicographical works. Preliminary results indicate a phone error rate accuracy of 97.5%, with the possibility of increase