Focus of this paper is statistical machine translation (SMT) system developed at ITC-irst for the evaluation campaign of the 2005 International Workshop on Spoken Language Translation (IWSLT 2005). As far as restrictions on available resources are concerned, we took part in the Supplied Data track only. Moreover the system was developed to translate from manual transcriptions. We evaluate our system on three language directions: Chinese-to-English, Japanese-to- English and Arabic-to-English. The system realized to translate from Japanese manual transcriptions is also evaluated when translating from the (Japanese) ASR Irst-best output. After a presentation of the statistical model underlying the system and its architecture, experiments and results from the evaluation campaign are reported and discussed.
The ITC-irst SMT System for IWSLT-2005
Cattoni, Roldano;Bertoldi, Nicola;Cettolo, Mauro;Federico, Marcello
2005-01-01
Abstract
Focus of this paper is statistical machine translation (SMT) system developed at ITC-irst for the evaluation campaign of the 2005 International Workshop on Spoken Language Translation (IWSLT 2005). As far as restrictions on available resources are concerned, we took part in the Supplied Data track only. Moreover the system was developed to translate from manual transcriptions. We evaluate our system on three language directions: Chinese-to-English, Japanese-to- English and Arabic-to-English. The system realized to translate from Japanese manual transcriptions is also evaluated when translating from the (Japanese) ASR Irst-best output. After a presentation of the statistical model underlying the system and its architecture, experiments and results from the evaluation campaign are reported and discussed.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
