Inductive Logic Programming for Corpus-Based Acquisition of Semantic Lexicons

J In: Proceedings of CoNLL-2000 and LLL-2000, pages 199-208, Lisbon, Portugal, 2000. Inductive Logic Programming for Corpus-Based Acquisition of Semantic Lexicons Pascale S4billot IRISA - Campus de Beaulieu - 35042 Rennes cedex - France sebillot@irisa, fr Pierrette Bouillon TIM/ISSCO - ETI - Universit4 de Gen~ve - 40 Bvd du Pont-d'Arve - CH-1205 Geneva- Switzerland Pierrette. Bouillon@issco. unige, ch C4cile Fabre ERSS - Universit@ de Toulouse II - 5 all@es A. Machado - 31058 Toulouse cedex - France cfabre@univ-tlse2, fr Abstract In this paper, we propose an Inductive Logic Programming learning method which aims at automatically extracting special Noun-Verb (N- V) pairs from a corpus in order to build up semantic lexicons based on Pustejovsky's Gen- erative Lexicon (GL) principles (Pustejovsky, 1995). In one of the components of this lex- ical model, called the qualia structure, words are described in terms of semantic roles. For example, the relic role indicates the purpose or function of an item (cut for knife), the agen- tive role its creation mode (build for house), etc. The qualia structure of a noun is mainly made up of verbal associations, encoding rela- tional information. The Inductive Logic Pro- gramming learning method that we have devel- oped enables us to automatically extract from a corpus N-V pairs whose elements axe linked by one of the semantic relations defined in the qualia structure in GL, and to distinguish them, in terms of surrounding categorial context from N-V pairs also present in sentences of the corpus but not relevant. This method has been theoret- ically and empirically validated, on a technical corpus. The N-V pairs that have been extracted will further be used in information retrieval ap- plications for index expansion 1. 1This works is funded by the Agence universi- taire de la Francophonie (AUF) (Action de recherche partag4e "Acquisition automatique d'dldments du Lex- Keywords: Lexicon learning, Generative Lexicon, Inductive Logic Programming, Infor- mation indexing. 1 Introduction Information retrieval (IR) systems aim at pro- viding a user who asks a query to a database of documents with the most relevant texts. The quality of these systems is usually measured with the help of two criteria: the recall rate, which corresponds to the proportion of relevant answers that have been given by the system compared to the total number of relevant an- swers in the database, and the precision rate, which denotes the proportion of relevant an- swers that are present among the given answers. In these IR systems, texts and queries are usually represented by indexes, that is, a col- lection of some of the words that they contain. The quality of the systems therefore highly de- pends on the type of indexing language that has been chosen. Two kinds of indexes exist: sim- ple indexes, which correspond to simple nouns (N), verbs (V) and/or adjectives (A) that oc- cur in a text or a query 2, and complex indexes, which correspond to the compounds (for exam- ple, NN compounds) present in the document or ique Gdndratif pour amdliorer les …

Content truncated to 500 words