ViPEr: A Lexicon-Grammar of European Portuguese Verbs

International audienceThis paper presents the current state of ViPEr, the Lexicon-Grammar of European Portuguese verbs, a database with distributional, syntactic and semantic properties of the most frequently occurring verbs. The classification follows the theoretical framework of the Lexicon-Grammar. The paper presents the main linguistic criteria that were adopted in the classification of 5052 frequently occurring verbs, which yield 6,059 different constructions or word senses. The paper concludes with some preliminary results on the application of ViPEr to texts and plans for future work. 1. Introduction The use of reliable, large-coverage language resources is key to the performance of many Natural Language Processing (NLP) systems. To our knowledge, while some syntactic descriptions of European Portuguese verbs exist, most have not been made publicly available to the NLP community or just consist in human-oriented dictionaries, not having been built for computational processing (FERNANDES 2008, BORBA 1991; BUSSE 1994). On the other hand, partial linguistic studies have produced throughout the last three decades, with major efforts in the late 90s (OLIVEIRA 1981; NASCIMENTO 1997; RODRIGUES 1997), but little, if any, use was made by the NLP community of such data, and very little effort has been addressed to validate or test those, mostly introspective-sourced, and theoretically-oriented, linguistic descriptions. A recent attempt in that direction is that of GOMES (2011), that highlighted the difficulties of the task. For this project, a practical approach to the lexicon was adopted. For many NLP tasks, but specially for any task where a fine-grained semantic distinction in required of ambiguous lexical forms, being able to identify the meaning of the verb (and of the surrounding elements as well) can be facilitated by the knowledge of the syntactic and semantic constraints the verb imposes on the lexical fulfillment of its argument positions. In particular, the number of verb arguments; their structural and distributional type; the prepositions the verb selects to introduce its essential complements; the main shape-changes that these structures can undergo; and other relevant linguistic information; besides its intrinsic linguistic interest, all this data can be put to use to improve parsing strategies, word-sense disambiguation, question-answer systems, computer-assisted language learning systems, among other applications. Above all, an inventory of basic word senses and their corresponding structures is necessary, and this is the aim of the ViPEr project. This paper presents the current state of the Lexicon-Grammar of European Portuguese verbs. The classification of European Portuguese verb constructions is largely based on the methodology presented by M. GROSS (1975, 1981, 1996) and his collaborators under the Lexicon-Grammar theoretical framework (see LAMIROY (1998) for an overview). The classification proper is directly inspired in the synthesis of LECLÈRE (2002). In the following, the main linguistic criteria that were adopted in the classification are presented. For lack of space, only the most salient classes and properties are presented. The paper concludes with preliminary results from applying this new resource to real texts, and prepares future work