Key points are not available for this paper at this time.
This paper describes the University of Ed-inburgh’s (UEDIN) phrase-based submis-sions to the translation and medical trans-lation shared tasks of the 2014 Work-shop on Statistical Machine Translation (WMT). We participated in all language pairs. We have improved upon our 2013 system by i) using generalized represen-tations, specifically automatic word clus-ters for translations out of English, ii) us-ing unsupervised character-based models to translate unknown words in Russian-English and Hindi-English pairs, iii) syn-thesizing Hindi data from closely-related Urdu data, and iv) building huge language on the common crawl corpus. 1 Translation Task Our baseline systems are based on the setup de-scribed in (Durrani et al., 2013b) that we used for the Eighth Workshop on Statistical Machine Translation (Bojar et al., 2013). The notable fea-tures of these systems are described in the follow-ing section. The experiments that we carried out for this year’s translation task are described in the following sections. 1.1 Baseline We trained our systems with the following set-tings: a maximum sentence length of 80, grow-diag-final-and symmetrization of GIZA++ align-ments, an interpolated Kneser-Ney smoothed 5-gram language model with KenLM (Heafield, 2011) used at runtime, hierarchical lexicalized re-ordering (Galley and Manning, 2008), a lexically-driven 5-gram operation sequence model (OSM) (Durrani et al., 2013a) with 4 count-based sup-portive features, sparse domain indicator, phrase length, and count bin features (Blunsom and Os-borne, 2008; Chiang et al., 2009), a distortion limit of 6, maximum phrase-length of 5, 100-best trans-
Durrani et al. (Wed,) studied this question.