VerbLexPor: um recurso léxico com anotação de papéis semânticos para o português
Abstract
Semantic role labeling offers vital information for both Linguistics and Natural Language Processing tasks. In this article, we present a lexical resource for Portuguese annotated with semantic roles: VerbLexPor. The resource is a database with verbs and sentences extracted from both a domain specific corpus and a non-specialized generic one. Annotation was manually carried out by a linguist using VerbNet-like semantic roles. The resource has more than 6 thousand annotated sentences and 15 thousand annotated arguments, and is available for download as XML or SQL files. The paper also describes a comparative analysis between the two corpora, showing that the distribution of semantic roles in a general domain is different from that in specific domain.
