WSD for n-best reranking and local language modeling in SMT
Marianna Apidianaki, Guillaume Wisniewski, Artem Sokolov, Aurélien Max, François Yvon · 2013
We integrate semantic information at two stages of the translation process of a state-ofthe-art SMT system. A Word Sense Disambiguation (WSD) classifier produces a probability distribution over the translation candidates of source words which is exploited in two ways. First, the probabilities serve to rerank a list of n-best translations produced by the system. Second, the WSD predictions are used to build a supplementary language model for each sentence, aimed to favor translations that seem more adequate in this specific sentential context. Both approaches lead to significant improvements in translation performance, highlighting the usefulness of source side disambiguation for SMT. 1