Exploring the Vector Space Model for Finding Verb Synonyms in Portuguese

We explore the performance of the Vector Space Model (VSM) in finding verb synonyms in Portuguese by analyzing the impact of three operating parameters: (i) the weighting function, (ii) the context window used for automatically extracting features, and (iii) the minimum number of vector features. We rely on distributional statistics taken from a large n-gram database to build feature vectors, using minimal linguistic pre-processing. Automatic evaluation of synonym candidates using gold-standard information from the OpenOffice and Wiktionary thesaurus shows that low frequency features carry most information regarding verb similarity, and that a [0, +2] window carries more information than a [-2, 0] window. We show that satisfactory precision levels require vectors with 50 or more non-nil components. Manual evaluation over a set of declarative verbs and psychological verbs show that VSM-based approaches achieve good precision in finding verb synonyms for Portuguese, even when using minimal linguistic knowledge. This lead us to proposing a performance baseline for this task.