Dynamic Topic Adaptation for SMT using Distributional Profiles

Proceedings of the Ninth Workshop on Statistical Machine Translation, pages 445–456, Baltimore, Maryland USA, June 26–27, 2014. c⃝2014 Association for Computational Linguistics Dynamic Topic Adaptation for SMT using Distributional Profiles Eva Hasler1 Barry Haddow1 Philipp Koehn1,2 1School of Informatics, University of Edinburgh 2Center for Language and Speech Processing, Johns Hopkins University e.hasler@ed.ac.uk, {bhaddow,pkoehn}@inf.ed.ac.uk Abstract Despite its potential to improve lexical selection, most state-of-the-art machine translation systems take only minimal con- textual information into account. We cap- ture context with a topic model over dis- tributional profiles built from the context words of each translation unit. Topic dis- tributions are inferred for each transla- tion unit and used to adapt the translation model dynamically to a given test context by measuring their similarity. We show that combining information from both lo- cal and global test contexts helps to im- prove lexical selection and outperforms a baseline system by up to 1.15 BLEU. We test our topic-adapted model on a diverse data set containing documents from three different domains and achieve competitive performance in comparison with two su- pervised domain-adapted systems. 1 Introduction The task of lexical selection plays an important role in statistical machine translation (SMT). It strongly depends on context and is particularly dif- ficult when the domain of a test document is un- known, for example when translating web doc- uments from diverse sources. Selecting transla- tions of words or phrases that preserve the sense of the source words is closely related to the field of word sense disambiguation (WSD), which has been studied extensively in the past. Most approaches to WSD model context at the sentence level and do not take the wider context of a word into account. Some of the ideas from the field of WSD have been adapted for machine translation (Carpuat and Wu, 2007b; Carpuat and Wu, 2007a; Chan et al., 2007). For example, Carpuat and Wu (2007a) extend word sense dis- ambiguation to phrase sense disambiguation and show improved performance due to the better fit with multiple possible segmentations in a phrase- based system. Carpuat (2009) test the “one sense per discourse” hypothesis (Gale et al., 1992) for MT and find that enforcing it as a constraint at the document level could potentially improve transla- tion quality. Our goal is to make correct lexical choices in a given context without explicitly en- forcing translation consistency. More recent work in SMT uses latent repre- sentations of the document context to dynam- ically adapt the translation model with either monolingual topic models (Eidelman et al., 2012; Hewavitharana et al., 2013) or bilingual topic models (Hasler et al., 2014), thereby allowing the translation system to disambiguate source phrases using document context. Eidelman et al. (2012) also apply a topic model to each test sentence and find that sentence context is sufficient for pick- ing good translations, but they do not attempt to combine sentence and document level informa- tion. Sentence-level topic adaptation for SMT has also been employed by Hasler et al. (2012). Other approaches to …

Content truncated to 500 words