Phrasenet: towards context sensitive lexical semantics

PhraseNet: Towards Context Sensitive Lexical Semantics∗ Xin Li†, Dan Roth†, Yuancheng Tu‡ Dept. of Computer Science† Dept. of Linguistics‡ University of Illinois at Urbana-Champaign {xli1,danr,ytu}@uiuc.edu Abstract This paper introduces PhraseNet, a context- sensitive lexical semantic knowledge base sys- tem. Based on the supposition that seman- tic proximity is not simply a relation between two words in isolation, but rather a relation between them in their context, English nouns and verbs, along with contexts they appear in, are organized in PhraseNet into Consets; Con- sets capture the underlying lexical concept, and are connected with several semantic relations that respect contextually sensitive lexical infor- mation. PhraseNet makes use of WordNet as an important knowledge source. It enhances a WordNet synset with its contextual informa- tion and refines its relational structure by main- taining only those relations that respect con- textual constraints. The contextual informa- tion allows for supporting more functionali- ties compared with those of WordNet. Nat- ural language researchers as well as linguists and language learners can gain from accessing PhraseNet with a word token and its context, to retrieve relevant semantic information. We describe the design and construction of PhraseNet and give preliminary experimental evidence to its usefulness for NLP researches. 1 Introduction Progress in natural language understanding research ne- cessitates significant progress in lexical semantics and the development of lexical semantics resources. In a broad range of natural language applications, from ∗Research supported by NSF grants IIS-99-84168, ITR-IIS-00-85836 and an ONR MURI award. Names of authors are listed alphabetically. prepositional phrase attachment (Pantel and Lin, 2000; Stetina and Nagao, 1997), co-reference resolution (Ng and Cardie, 2002) to text summarization (Saggion and Lapalme, 2002), semantic information is a necessary component in the inference, by providing a level of ab- straction that is necessary for robust decisions. Inducing that the prepositional phrase in “They ate a cake with a fork” has the same grammatical function as that in “They ate a cake with a spoon”, for example, depends on the knowledge that “cutlery” and “tableware” are the hypernyms of both “fork” and “spoon”. However, the noun “fork” has five senses listed in WordNet and each of them has several different hypernyms. Choosing the correct one is a con- text sensitive decision. WordNet (Fellbaum, 1998), a manually constructed lexical reference system provides a lexical database along with semantic relations among the lexemes of English and is widely used in NLP tasks today. However, Word- Net is organized at the word level, and at this level, En- glish suffers ambiguities. Stand-alone words may have several meanings and take on relations (e.g., hypernyms, hyponyms) that depend on their meanings. Consequently, there are very few success stories of automatically us- ing WordNet in natural language applications. In many cases, reported (and unreported) problems are due to the fact that WordNet enumerates all the senses of polyse- mous words; attempts to use this resource automatically often result in noisy and non-uniform information (Brill and Resnik, 1994; Krymolowski and Roth, 1998). PhraseNet is designed based on the …

Content truncated to 500 words