ERCIM News No.26 - July 1996 - INESC

INESC's Speech Processing Group

by Isabel Trancoso

The Speech Processing group at INESC is actively involved in the fields of coding (specially high-quality medium-to-low bit rate coding), recognition and synthesis.

This activity has been internationally recognized through intense involvement in European Projects and close cooperation with prestigious institutions such as AT&T Bell Laboratories (New Jersey, U.S.A.). At a national level, the group has a formal cooperation agreement with the Centre of Linguistics of the University of Lisbon (CLUL), having jointly developed the first Text-to-Speech synthesizer build from scratch for European Portuguese (DIXI). The cooperation with the Neural Network and the Natural Language Processing groups of INESC is also quite active.

Since the Portuguese integration in the EC, the group has participated very intensively in European projects, focusing on the ESPRIT Programme (ESPRIT-II SUNSTAR - Integration and design of speech understanding interfaces and ESPRIT-III SAMA - Speech Technology Assessment in Multilingual Applications) and the LRE (ONOMASTICA - Multi-Language Pronounciation of Proper Names and Place Names and RELATOR - a European Network of Repositaries for Linguistic Resources). It has also participated in EUREKA (High Quality Coding) and COST (Applications of Digital Signal Processing to Communications). The group also developed a national project to design a prototype of a recognition/synthesis system in the Portuguese language applied to database queries over the telephone network.

The Speech Processing group is currently participating in several projects, namely the ESPRIT Network in Language and Speech, where INESC acts as one of the managing nodes, being involved in the Linguistic Resources task group and, related to this, the HCM-ELSNET Project on Phrase Level Phonology and Dialogue and Discourse. Other European projects include the LRE Project Speechdat and the Telematics Project VODIS - Advanced Speech Technologies for Voice Operated Driver Information Systems. Nationally, the group is active in the projects BDFALA - Spoken Corpus of European Portuguese, and EDIFALA - Vocal Aids for the Motor and Speech Handicapped: integration of Text-to-Speech Synthesis in a Text Editor.

The group is also launching a new electronic journal on Language and Speech, devoted to papers from students, the 'Student Journal of Language and Speech'. The journal has the active support of ELSNET, ESCA (the European Speech Communication Association) and EACL (the European Chapter of the Association of Computational Linguists).

The new electronic journal is intended as a means for rapid dissemination of ideas and results from research students working within the fields of study and the geographical area covered by the three sponsoring institutions and has a training role for new researchers.

Given the wide coverage of the journal in terms of scientific areas, 9 very broad classes were selected: SRS - speech recognition and synthesis; NLUG - natural language understanding and generation; LI - language identification; IER - information extraction and retrieval; DMC - dialogue modelling and control; MLPP - models of language perception and production; SM - stochastic modelling; GCLD - grammars, corpora, lexicons1 and databases; SAS - system assessment strategies. The web address is:

Please contact:
Isabel Trancoso - INESC
Tel: +351 1 3100268

return to the contents page