Integrating WordNet and FrameNet using a Knowledge-based Word Sense Disambiguation Algorithm

International Conference RANLP 2009 - Borovets, Bulgaria, pages 208–213 Integrating WordNet and FrameNet using a knowledge-based Word Sense Disambiguation algorithm Egoitz Laparra and German Rigau IXA group. University of the Basque Country, Donostia {egoitz.laparra,german.rigau}@ehu.com Abstract This paper presents a novel automatic approach to partially integrate FrameNet and WordNet. In that way we expect to extend FrameNet coverage, to en- rich WordNet with frame semantic information and possibly to extend FrameNet to languages other than English. The method uses a knowledge-based Word Sense Disambiguation algorithm for linking FrameNet lexical units to WordNet synsets. Specifically, we ex- ploit a graph-based Word Sense Disambiguation algo- rithm that uses a large-scale knowledge-base derived from WordNet. We have developed and tested four additional versions of this algorithm showing a sub- stantial improvement over previous results. 1 Introduction Predicate models such as FrameNet [6], VerbNet [16] or PropBank [23] are core resources in most advanced NLP tasks, such as Question Answering, Textual Entailment or Information Extraction. Most of the systems with Natural Language Understanding capabilities require a large and precise amount of semantic knowledge at the predicate- argument level. This type of knowledge allows to iden- tify the underlying typical participants of a particular event independently of its realization in the text. Thus, us- ing these models, different linguistic phenomena express- ing the same event, such as active/passive transformations, verb alternations and nominalizations can be harmonized into a common semantic representation. In fact, lately, several systems have been developed for shallow seman- tic parsing and semantic role labeling using these resources [11], [26], [14]. However, building large and rich enough predicate mod- els for broad–coverage semantic processing takes a great deal of expensive manual effort involving large research groups during long periods of development. Thus, the cov- erage of currently available predicate-argument resources is still unsatisfactory. For example, [7] or [25] indicate the limited coverage of FrameNet as one of the main prob- lems of this resource. In fact, FrameNet1.3 covers around 10,000 lexical-units while for instance, WordNet3.0 con- tains more than 150,000 words. Furthermore, the same ef- fort should be invested for each different language [27]. Following the line of previous works [26], [7], [15], [24], [8], [29], we empirically study a novel approach to partially integrate FrameNet [6] and WordNet [12]. The method re- lies on the use of a knowledge-based Word Sense Disam- biguation (WSD) algorithm that uses a large-scale graph of concepts derived from WordNet [12] and eXtented Word- Net [19]. The WSD algorithm is applied to semantically coherent groupings of words belonging to the same frame. In that way we expect to extend the coverage of FrameNet (by including from WordNet closely related concepts), to enrich WordNet with frame semantic information (by port- ing frame information to WordNet) and possibly to extend FrameNet to languages other than English (by exploiting local wordnets aligned to the English WordNet). WordNet1 [12] (hereinafter WN) is by far the most widely-used knowledge base. In fact, WN is being used world-wide for anchoring different types of …

Content truncated to 500 words