A non supervis ed method for sentiment polarity detection on reported speech from news

This papers shows the results obtained by a non supervised method in the task of sentiment polarity detection on news quotations (reported speech) from the 2010 JRC 1590 quotes corpus. This method, which obtains a subgraph of WordNet by applying Page Rank over it, weights such a graph with correspond- ing SentiWordNet polarity scores, offering a non-supervised solution comparable to state-of-the-art techniques.