The IWSLT 2014 Evaluation Campaign includes the MT track on TED Talks. In this edition, the official language pairs are five:
from English to French
from English to German
from German to English
from English to Italian
from Italian to English
Optional tasks are proposed with English paired in both directions with other twelve languages:
from/to English to/from Arabic, Spanish, Farsi, Hebrew, Dutch, Polish, Portuguese-Brazil, Romanian, Russian, Slovenian, Turkish and Chinese
Submitted runs on additional pairs will be evaluated as well, in the hope to stimulate the MT community to evaluate systems on common benchmarks and to share achievements on challenging translation tasks.
For each language pair, training and development sets are linked to the corresponding entry of the table below: by clicking, an archive will be downloaded which contains the sets and a README file. Numbers in the table refer to millions of units (untokenized words) of the target side of parallel training data.
Reference performance of baseline systems are available here.
Linguistic preprocessors for normalize/cleaning Arabic, Farsi and Hebrew texts are available here; they were used for the development and the scoring of baseline systems and will be employed for the official evaluation of submissions.
If you use this corpus in your work, please cite the paper: