A Robust Approach to Aligning Heterogeneous Lexical Resources

Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, pages 468–478, Baltimore, Maryland, USA, June 23-25 2014. c⃝2014 Association for Computational Linguistics A Robust Approach to Aligning Heterogeneous Lexical Resources Mohammad Taher Pilehvar and Roberto Navigli Department of Computer Science Sapienza University of Rome {pilehvar,navigli}@di.uniroma1.it Abstract Lexical resource alignment has been an active field of research over the last decade. However, prior methods for align- ing lexical resources have been either spe- cific to a particular pair of resources, or heavily dependent on the availability of hand-crafted alignment data for the pair of resources to be aligned. Here we present a unified approach that can be applied to an arbitrary pair of lexical resources, includ- ing machine-readable dictionaries with no network structure. Our approach leverages a similarity measure that enables the struc- tural comparison of senses across lexical resources, achieving state-of-the-art per- formance on the task of aligning WordNet to three different collaborative resources: Wikipedia, Wiktionary and OmegaWiki. 1 Introduction Lexical resources are repositories of machine- readable knowledge that can be used in virtually any Natural Language Processing task. Notable examples are WordNet, Wikipedia and, more re- cently, collaboratively-curated resources such as OmegaWiki and Wiktionary (Hovy et al., 2013). On the one hand, these resources are heteroge- neous in design, structure and content, but, on the other hand, they often provide complemen- tary knowledge which we would like to see inte- grated. Given the large scale this intrinsic issue can only be addressed automatically, by means of lexical resource alignment algorithms. Owing to its ability to bring together features like multilin- guality and increasing coverage, over the past few years resource alignment has proven beneficial to a wide spectrum of tasks, such as Semantic Pars- ing (Shi and Mihalcea, 2005), Semantic Role La- beling (Palmer et al., 2010), and Word Sense Dis- ambiguation (Navigli and Ponzetto, 2012). Nevertheless, when it comes to aligning textual definitions in different resources, the lexical ap- proach (Ruiz-Casado et al., 2005; de Melo and Weikum, 2010; Henrich et al., 2011) falls short because of the potential use of totally different wordings to define the same concept. Deeper ap- proaches leverage semantic similarity to go be- yond the surface realization of definitions (Nav- igli, 2006; Meyer and Gurevych, 2011; Niemann and Gurevych, 2011). While providing good re- sults in general, these approaches fail when the definitions of a given word are not of adequate quality and expressiveness to be distinguishable from one another. When a lexical resource can be viewed as a semantic graph, as with WordNet or Wikipedia, this limit can be overcome by means of alignment algorithms that exploit the network structure to determine the similarity of concept pairs. However, not all lexical resources provide explicit semantic relations between concepts and, hence, machine-readable dictionaries like Wik- tionary have first to be transformed into semantic graphs before such graph-based approaches can be applied to them. To do this, recent work has pro- posed graph construction by monosemous linking, where a concept is …

Content truncated to 500 words