The time invested in dealing with email has significantly increased for most computer users. Speech technologies, such as recognition and synthesis, can he used to improve the productivity of user or even accessibility for those who have special needs. This work describes the development of a plugin for the Mozilla Thunderbid that allows not only controlling the software via voice commands, but also to convert text into speech. The project relies on the free software philosophy and benefits from recent advances in speech processing research targeting Brazilian Portuguese. One of the contributions of the work is to significantly improve the system recognition rate by supporting commandand-control grammars. Previous versions of the adopted programming interface were limited to N-gram language models, which are more adequate to applications such as dictation. 1. Introdução Um importante passo para a melhoria da interação dos usuários com aplicativos de computador é a possibilidade de comunicação através de linguagem natural falada. Esse desafio é considerado por muitos como um dos mais importantes da computação moderna. Para isso dois importantes processos são requeridos: primeiramente o processo pelo qual o computador interpreta o que o usuário fala, automatic speech recognition (ASR) [Taylor 2009], e o segundo, igualmente importante, é a produção de linguagem natural pela
[1]
Patrick Silva,et al.
An Open-Source Speech Recognizer for Brazilian Portuguese with a Windows Programming Interface
,
2010,
PROPOR.
[2]
Alex Acero,et al.
Spoken Language Processing
,
2001
.
[3]
Stephen R. Barley,et al.
E-mail as a Source and Symbol of Stress
,
2011,
Organ. Sci..
[4]
Kiyohiro Shikano,et al.
Julius - an open source real-time large vocabulary recognition engine
,
2001,
INTERSPEECH.
[5]
Joseph P. Olive,et al.
Text-to-speech synthesis
,
1995,
AT&T Technical Journal.
[6]
John F. Canny,et al.
The Future of Human-Computer Interaction
,
2006,
ACM Queue.
[7]
Isabel Trancoso,et al.
Free tools and resources for Brazilian Portuguese speech recognition
,
2011,
Journal of the Brazilian Computer Society.