Sew-Embed at SemEval-2017 Task 2: Language-Independent Concept Representations from a Semantically Enriched Wikipedia

Proceedings of the 11th International Workshop on Semantic Evaluations (SemEval-2017), pages 261–266, Vancouver, Canada, August 3 - 4, 2017. c⃝2017 Association for Computational Linguistics SEW-EMBED at SemEval-2017 Task 2: Language-Independent Concept Representations from a Semantically Enriched Wikipedia Claudio Delli Bovi and Alessandro Raganato Department of Computer Science Sapienza University of Rome {dellibovi,raganato}@di.uniroma1.it Abstract This paper describes SEW-EMBED, our language-independent approach to multi- lingual and cross-lingual semantic word similarity as part of the SemEval-2017 Task 2. We leverage the Wikipedia- based concept representations developed by Raganato et al. (2016), and propose an embedded augmentation of their ex- plicit high-dimensional vectors, which we obtain by plugging in an arbitrary word (or sense) embedding representation, and computing a weighted average in the con- tinuous vector space. We evaluate SEW- EMBED with two different off-the-shelf embedding representations, and report their performances across all monolin- gual and cross-lingual benchmarks avail- able for the task. Despite its simplic- ity, especially compared with supervised or overly tuned approaches, SEW-EMBED achieves competitive results in the cross- lingual setting (3rd best result in the global ranking of subtask 2, score 0.56). 1 Introduction Semantic similarity is a well established research area of Natural Language Processing, concerned with measuring the extent to which two linguistic items are similar (Budanitsky and Hirst, 2006). In particular, word similarity is nowadays a widely used evaluation benchmark for word and sense representations (Turney and Pantel, 2010). While many classical approaches to word sim- ilarity have been limited to the English lan- guage (Gabrilovich and Markovitch, 2007; Mi- halcea, 2007; Pilehvar et al., 2013; Baroni et al., 2014), a growing interest for multilingual and cross-lingual models is emerging (Hassan and Mi- halcea, 2011; Camacho Collados et al., 2016) and it is accompanied by the development of multilin- gual benchmarks (Gurevych, 2005; Granada et al., 2014; Camacho Collados et al., 2015). In this respect Wikipedia, as one of the most popular semi-structured resources in the field (Hovy et al., 2013), provides a convenient bridge to multilinguality, with several million inter-language links among articles refferring to the same concept or entity. In fact, a number of successful approaches to semantic similarity make explicit use of Wikipedia, from ESA (Gabrilovich and Markovitch, 2007) to NASARI (Camacho Col- lados et al., 2016). Others, like SENSEMBED (Ia- cobacci et al., 2015), report state-of-the-art re- sults when trained on an automatically disam- biguated version of a Wikipedia dump. Regard- less of whether Wikipedia is seen as a multi- lingual semantic network of concepts and enti- ties or as a sense-annotated corpus, hyperlinks (inter-page links) constitute its key structural prop- erty: in light of this, Raganato et al. (2016) ad- dressed the sparsity problem of original hyper- links and developed SEW1, a semantically en- riched Wikipedia where the overall number of linked mentions has been more than tripled by solely exploiting the structure of Wikipedia itself and the wide-coverage sense inventory of Babel- Net (Navigli and Ponzetto, 2012)2. In addition to building the corpus, the authors used SEW’s sense annotations to construct …

Content truncated to 500 words