No abstract

This paper describes the joined submission of LIMSI and KIT to the Shared Translation Task for the German-toEnglish direction. The system consists of a phrase-based translation system using a pre-reordering approach. The baseline system already includes several models like conventional language models on different word factors and a discriminative word lexicon. This system is used to generate a k-best list. In a second step, the list is reranked using SOUL language and translation models (Le et al., 2011).Originally, SOUL translation models were applied to n-gram-based translation systems that use tuples as translation units instead of phrase pairs. In this article, we describe their integration into the KIT phrase-based system. Experimental results show that their use can yield significant improvements in terms of BLEU score.

This paper describes LIMSI's submission to the first medical translation task at WMT'14. We report results for EnglishFrench on the subtask of sentence translation from summaries of medical articles.Our main submission uses a combination of NCODE (n-gram-based) and MOSES (phrase-based) output and continuous-space language models used in a post-processing step for each system. Other characteristics of our submission include: the use of sampling for building MOSES' phrase table; the implementation of the vector space model proposed by Chen et al. (2013); adaptation of the POStagger used by NCODE to the medical domain; and a report of error analysis based on the typology of Vilar et al. (2006).

This paper explores a new discriminative training procedure for continuousspace translation models (CTMs) which correlates better with translation quality than conventional training methods. The core of the method lays in the definition of a novel objective function which enables us to effectively integrate the CTM with the rest of the translation system through N-best rescoring. Using a fixed architecture, where we iteratively retrain the CTM parameters and the log-linear coefficients, we compare various ways to define and combine training criteria for each of these steps, drawing inspirations both from max-margin and learning-to-rank techniques. We experimentally show that a recently introduced loss function, which combines these two techniques, outperforms several objective functions from the literature. We also show that ensuring the consistency of the losses used to train these two sets of parameters is beneficial to the overall performance.