Assigning Polarity Automatically to the Synsets of a Wordnet-like Resource

This article describes work towards the automatic creation of a conceptual polarity lexicon for Portuguese. For this purpose, we take advantage of a polarity lexicon based on single lemmas to assign polarities to the synsets of a wordnet-like resource. We assume that each synset has the polarity of the majority of its lemmas, given by the initial lexicon. After that, polarity is propagated to other synsets, through different types of semantic relations. The relation types used were selected after manual evaluation. The main result of this work is a lexicon with more than 10,000 synsets with an assigned polarity, with accuracy of 70% or 79%, depending on the human evaluator. For Portuguese, this is the first synset-based polarity lexicon we are aware of. In addition to this contribution, the presented approach can be applied to create similar resources for other languages. 1998 ACM Subject Classification I.2.7 Natural Language Processing