TC-STAR Third Evaluation Campaign 2007

TC-STAR is an European integrated project focusing on Speech-to-Speech Translation (SST). To encourage significant advances in all SST technologies, annual competitive evaluations are organized. Automatic Speech Recognition (ASR), Spoken Language Translation (SLT) and Text-To-Speech (TTS) are evaluated independently and within an end-to-end system. The project targets a selection of unconstrained conversational speech domains—speeches and broadcast news—and three languages: European English, European Spanish, and Mandarin Chinese.

The second evaluation campaign took place in March 2006 and was concluded by an international workshop. TC-STAR welcomes external participants in its 3rd evaluation of January-February 2007. This participation is free of charge.

The TC-STAR 2007 evaluation campaign will consider:

SLT in the following directions
a) Chinese-to-English (Broadcast News)
b) Spanish-to-English (European Parliament plenary speeches; Cortes Spanish Parliament speeches)
c) English-to-Spanish (European Parliament plenary speeches)

The SLT evaluation of European Parliament plenary speeches will cover three types of input:
1. ASR output (single best sentence, and more: N-best, word lattices)
2. verbatim transcription,
3. final text edition.

ASR in the following languages
a) English (European Parliament plenary speeches)
b) Spanish (European Parliament plenary speeches; Cortes Spanish Parliament speeches)
c) Mandarin Chinese (Broadcast News)

TTS in Chinese, English, and Spanish under the following conditions:
a) Complete system: participants use their own training data
b) Voice conversion intralingual and crosslingual, expressive speech: data provided by TC-STAR
c) Component evaluation

For ASR and SLT, training data will be made available by the TC-STAR project for English and Spanish and can be purchased at LDC for Chinese. Development data will be provided by the TC-STAR project. Legal issues regarding the data will be detailed in the 2nd Call For Participation.

The participation is free of charge and training/development/evaluation data from TC-STAR will also be made freely available to participants."

All participants will be given the opportunity to present and discuss their results in the 3rd TC-STAR evaluation workshop in Aachen in March 2007.

Tentative schedule:
Registration: October 2006 (early expression of interest is welcome)
ASR evaluation: from 21st of January to 28th of January 2007
SLT evaluation: from 31st of January to 7th February 2007
TTS evaluation: from 9th of February to15th of February 2007

Workshop: March 28-30, 2007 in Aachen, Germany

Contact: Djamel Mostefa (ELDA)
e-mail: mostefa at
tel. +33 1 43 13 33 33