Watset: Automatic Induction of Synsets from a Graph of Synonyms

Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1579–1590 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1145 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics, pages 1579–1590 Vancouver, Canada, July 30 - August 4, 2017. c⃝2017 Association for Computational Linguistics https://doi.org/10.18653/v1/P17-1145 Watset: Automatic Induction of Synsets from a Graph of Synonyms Dmitry Ustalov†∗, Alexander Panchenko‡, and Chris Biemann‡ †Institute of Natural Sciences and Mathematics, Ural Federal University, Russia ∗Krasovskii Institute of Mathematics and Mechanics, Russia ‡Language Technology Group, Department of Informatics, Universit¨at Hamburg, Germany dmitry.ustalov@urfu.ru {panchenko,biemann}@informatik.uni-hamburg.de Abstract This paper presents a new graph-based approach that induces synsets using syn- onymy dictionaries and word embeddings. First, we build a weighted graph of syn- onyms extracted from commonly available resources, such as Wiktionary. Second, we apply word sense induction to deal with ambiguous words. Finally, we clus- ter the disambiguated version of the am- biguous input graph into synsets. Our meta-clustering approach lets us use an efficient hard clustering algorithm to per- form a fuzzy clustering of the graph. De- spite its simplicity, our approach shows excellent results, outperforming five com- petitive state-of-the-art methods in terms of F-score on three gold standard datasets for English and Russian derived from large-scale manually constructed lexical resources. 1 Introduction A synset is a set of mutual synonyms, which can be represented as a clique graph where nodes are words and edges are synonymy relations. Synsets represent word senses and are building blocks of WordNet (Miller, 1995) and similar resources such as thesauri and lexical ontologies. These re- sources are crucial for many natural language pro- cessing applications that require common sense reasoning, such as information retrieval (Gong et al., 2005) and question answering (Kwok et al., 2001; Zhou et al., 2013). However, for most lan- guages, no manually-constructed resource is avail- able that is comparable to the English WordNet in terms of coverage and quality. For instance, Kiselev et al. (2015) present a comparative anal- ysis of lexical resources available for the Russian language concluding that there is no resource com- pared to WordNet in terms of coverage and qual- ity for Russian. This lack of linguistic resources for many languages urges the development of new methods for automatic construction of WordNet- like resources. The automatic methods foster con- struction and use of the new lexical resources. Wikipedia1, Wiktionary2, OmegaWiki3 and other collaboratively-created resources contain a large amount of lexical semantic information— yet designed to be human-readable and not for- mally structured. While semantic relations can be automatically extracted using tools such as DKPro JWKTL4 and Wikokit5, words in these re- lations are not disambiguated. For instance, the synonymy pairs (bank, streambank) and (bank, banking company) will be connected via the word “bank”, while they refer to the different senses. This problem stems from the fact that articles in Wiktionary and similar resources list undis- ambiguated synonyms. They are easy to disam- biguate for humans while reading a dictionary ar- …

Content truncated to 500 words