Towards a More Efficient Development of Statistical Machine Translation Systems (Vers un développement plus efficace des systèmes de traduction statistique : un peu de vert dans un monde de BLEU) [in French]

In this article, we show how using both on-demand alignment of parallel sentences and on-the-fly estimation of translation models can yield massive reduction (up to 93% in our experiments) in development time as compared to a state-of-the-art system, while offering an interesting tradeoff as regards translation quality under some configurations. We show in particular that the absence of a development set can be quickly compensated by immediately using translated documents. Mots-clés : traduction automatique statistique ; développement efficace ; temps de calcul.