INTERSPEECH 2004 - ICSLP
Multilingual spoken language corpora are indispensable for developing new speech-to-speech machine translation (S2SMT) technologies. This paper first discusses characteristics that corpora for S2SMT should have, then surveys existing corpora. Finally, it compares these corpora focusing on relations between collected data and collection scheme including instructions given to speakers.
Bibliographic reference. Kikui, Genichiro / Takezawa, Toshiyuki / Yamamoto, Seiichi (2004): "Multilingual corpora for speech-to-speech translation research", In INTERSPEECH-2004, 357-360.