Align, Disambiguate and Walk: A Unified Approach for Measuring Semantic Similarity

Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 1341–1351, Sofia, Bulgaria, August 4-9 2013. c⃝2013 Association for Computational Linguistics Align, Disambiguate and Walk: A Unified Approach for Measuring Semantic Similarity Mohammad Taher Pilehvar, David Jurgens and Roberto Navigli Department of Computer Science Sapienza University of Rome {pilehvar,jurgens,navigli}@di.uniroma1.it Abstract Semantic similarity is an essential com- ponent of many Natural Language Pro- cessing applications. However, prior meth- ods for computing semantic similarity of- ten operate at different levels, e.g., sin- gle words or entire documents, which re- quires adapting the method for each data type. We present a unified approach to se- mantic similarity that operates at multiple levels, all the way from comparing word senses to comparing text documents. Our method leverages a common probabilistic representation over word senses in order to compare different types of linguistic data. This unified representation shows state-of- the-art performance on three tasks: seman- tic textual similarity, word similarity, and word sense coarsening. 1 Introduction Semantic similarity is a core technique for many topics in Natural Language Processing such as Textual Entailment (Berant et al., 2012), Seman- tic Role Labeling (F¨urstenau and Lapata, 2012), and Question Answering (Surdeanu et al., 2011). For example, textual similarity enables relevant documents to be identified for information re- trieval (Hliaoutakis et al., 2006), while identifying similar words enables tasks such as paraphrasing (Glickman and Dagan, 2003), lexical substitution (McCarthy and Navigli, 2009), lexical simplifica- tion (Biran et al., 2011), and Web search result clustering (Di Marco and Navigli, 2013). Approaches to semantic similarity have often operated at separate levels: methods for word sim- ilarity are rarely applied to documents or even sin- gle sentences (Budanitsky and Hirst, 2006; Radin- sky et al., 2011; Halawi et al., 2012), while document-based similarity methods require more linguistic features, which often makes them in- applicable at the word or microtext level (Salton et al., 1975; Maguitman et al., 2005; Elsayed et al., 2008; Turney and Pantel, 2010). Despite the potential advantages, few approaches to semantic similarity operate at the sense level due to the chal- lenge in sense-tagging text (Navigli, 2009); for ex- ample, none of the top four systems in the recent SemEval-2012 task on textual similarity compared semantic representations that incorporated sense information (Agirre et al., 2012). We propose a unified approach to semantic sim- ilarity across multiple representation levels from senses to documents, which offers two signifi- cant advantages. First, the method is applicable independently of the input type, which enables meaningful similarity comparisons across differ- ent scales of text or lexical levels. Second, by op- erating at the sense level, a unified approach is able to identify the semantic similarities that exist in- dependently of the text’s lexical forms and any se- mantic ambiguity therein. For example, consider the sentences: t1. A manager fired the worker. t2. An employee was terminated from work by his boss. A surface-based approach would label the sen- tences as dissimilar due to the minimal lexical overlap. However, a sense-based …

Content truncated to 500 words