The KIT Translation Systems for IWSLT 2015
Thanh-Le Ha, Jan Niehues, Eunah Cho, Mohammed Mediani, Alex Waibel · 2015
In this paper, we present the KIT systems participating in the TED translation tasks of the IWSLT 2015 machine translation evaluation. We submitted phrase-based translation systems for three directions, namely English!German, German!English, and English!Vietnamese. For the official directions (English!German and German!English), we built systems both for the machine translation (MT) as well as the spoken language translation (SLT) tracks. This year we improved our systems’ performance over last year through n-best list rescoring using neural networkbased translation and language models and novel discriminative models based on different source-side features and classification methods. For the SLT tracks, we used a monolingual translation system to translate the lowercased ASR hypotheses with all punctuation stripped to truecased, punctuated output as a preprocessing step to our usual translation system. In addition to punctuation insertion, we also trained that system for sentence boundary insertion since the SLT’s data this year come with no sentence boundary.