Latent Semantic Word Sense Induction and Disambiguation

Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics, pages 1476–1485, Portland, Oregon, June 19-24, 2011. c⃝2011 Association for Computational Linguistics Latent Semantic Word Sense Induction and Disambiguation Tim Van de Cruys RCEAL University of Cambridge United Kingdom tv234@cam.ac.uk Marianna Apidianaki Alpage, INRIA & Univ Paris Diderot Sorbonne Paris Cit´e, UMRI-001 75013 Paris, France marianna.apidianaki@inria.fr Abstract In this paper, we present a unified model for the automatic induction of word senses from text, and the subsequent disambiguation of particular word instances using the automati- cally extracted sense inventory. The induction step and the disambiguation step are based on the same principle: words and contexts are mapped to a limited number of topical dimen- sions in a latent semantic word space. The in- tuition is that a particular sense is associated with a particular topic, so that different senses can be discriminated through their association with particular topical dimensions; in a similar vein, a particular instance of a word can be dis- ambiguated by determining its most important topical dimensions. The model is evaluated on the SEMEVAL-2010 word sense induction and disambiguation task, on which it reaches state- of-the-art results. 1 Introduction Word sense induction (WSI) is the task of automati- cally identifying the senses of words in texts, with- out the need for handcrafted resources or manually annotated data. The manual construction of a sense inventory is a tedious and time-consuming job, and the result is highly dependent on the annotators and the domain at hand. By applying an automatic proce- dure, we are able to only extract the senses that are objectively present in a particular corpus, and it al- lows for the sense inventory to be straightforwardly adapted to a new domain. Word sense disambiguation (WSD), on the other hand, is the closely related task of assigning a sense label to a particular instance of a word in context, using an existing sense inventory. The bulk of WSD algorithms up till now use pre-defined sense inven- tories (such as WordNet) that often contain fine- grained sense distinctions, which poses serious prob- lems for computational semantic processing (Ide and Wilks, 2007). Moreover, most WSD algorithms take a supervised approach, which requires a signifi- cant amount of manually annotated training data. The model presented here induces the senses of words in a fully unsupervised way, and subsequently uses the induced sense inventory for the unsuper- vised disambiguation of particular occurrences of words. The induction step and the disambiguation step are based on the same principle: words and contexts are mapped to a limited number of topical dimensions in a latent semantic word space. The key idea is that the model combines tight, synonym- like similarity (based on dependency relations) with broad, topical similarity (based on a large ‘bag of words’ context window). The intuition in this is that the dependency features can be disambiguated by the topical dimensions identified by the broad con- textual features; in a similar vein, a particular in- stance of a word can …

Content truncated to 500 words