Identification of bilingual terms from monolingual documents for statistical machine translation

The automatic translation of domain-specific documents is often a hard task for generic Statistical Machine Translation (SMT) systems, which are not able to correctly translate the large number of terms encountered in the text. In this paper, we address the problems of automatic identification of bilingual terminology using Wikipedia as a lexical resource, and its integration into an SMT system. The correct translation equivalent of the disambiguated term identified in the monolingual text is obtained by taking advantage of the multilingual versions of Wikipedia. This approach is compared to the bilingual terminology provided by the Terminology as a Service (TaaS) platform. The small amount of high quality domain-specific terms is passed to the SMT system using the XML markup and the Fill-Up model methods, which produced a relative translation improvement up to 13% BLEU score points. ; This publication has emanated from research supported in part by a research grant from Science Foundation Ireland (SFI) under Grant Number SFI/12/RC/2289 and by the European Union supported projects EuroSentiment (Grant No. 296277), LIDER (Grant No. 610782) and MateCat (ICT-2011.4.2-287688). ; non-peer-reviewed