Identifying Pronominal Verbs: Towards Automatic Disambiguation of the Clitic 'se' in Portuguese
Magali Sanches Duran, Carolina Scarton, Sandra Maria Aluísio, Carlos Ramisch · 2013
A challenging topic in Portuguese language processing is the multifunctional and ambiguous use of the clitic pronoun se, which impacts NLP tasks such as syntactic parsing, semantic role labeling and machine translation. Aiming to give a step forward towards the automatic disambiguation of se, our study focuses on the identification of pronominal verbs, which correspond to one of the six uses of se as a clitic pronoun, when se is considered a CONSTITU-TIVE PARTICLE of the verb lemma to which it is bound, as a multiword unit. Our strategy to identify such verbs is to analyze the results of a corpus search and to rule out all the other possible uses of se. This process evidenced the features needed in a computational lexicon to automatically perform the disambiguation task. The availability of the resulting lexicon of pronominal verbs on the web enables their inclusion in broader lexical resources, such as the Portuguese versions of Wordnet, Propbank and VerbNet. Moreover, it will allow the revision of parsers and dictionaries already in use. 1