| November 27, 2006 | |
| TC-STAR: a Speech to Speech Translation Project | |
| 13:30-14:30 →slides |
Dr. Gianni Lazzari (ITC-irst, Italy) |
|
Speech to speech translation is a challenging research problem. To simplify the task and make it somewhat more feasible, the first research project addressing this topic in the 1990's worked on restricted domains such as scheduling appointment.
TC-STAR (Technology and Corpora for Speech to Speech Translation) is the first joint research project addressing speech-to-speech translation in an unrestricted domain. TC-STAR works on European parliamentary speeches in native or non-native English and Spanish. The topics of the speeches include everything that is discussed in the European Parliament. Moreover the translation of Chinese broadcast news into Englis was chosen in order to link the project to available international benchmarks. My talk will give an overview of the TC-STAR project, the main achievements and the results of the last evaluation campaign. |
|
| November 28, 2006 | |
| Recent Results on MT Evaluation in the GALE Program | |
| 13:30-14:30 →slides |
Dr. Salim Roukos (IBM, US) |
| We will give an overview of the first year's evaluation results of the GALE program that is based on human post editing of the output of MT systems. A post-editor edits the MT system output untill the same meaning is conveyed as in the "Gold" reference. The Human Translation Error Rate (HTER) counts the number of edits performed by a post-editor normalized by the length of the "Gold" reference as the MT error metric. We will report on the sensitivity and stability of the new HTER metric for evaluating MT systems. We also compare the correlation of various automated metrics (BLEU, TER) to HTER. | |