SentiWordNet 3.0: An Enhanced Lexical Resource for Sentiment Analysis and Opinion Mining.

摘要: In this work we present SENTIWORDNET 3.0, a lexical resource explicitly devised for supporting sentiment classification and opinion mining applications. 3.0 is an improved version of 1.0, publicly available research purposes, now currently licensed to more than 300 groups used in variety projects worldwide. Both 1.0 are the result automatically annotating all WORDNET synsets according their degrees positivity, negativity, neutrality. differ (a) versions which they annotate (WORDNET 2.0 respectively), (b) algorithm WORDNET, includes (additionally previous semi-supervised learning step) random-walk step refining scores. We here discuss especially focussing on improvements concerning aspect that it embodies with respect 1.0. also report results evaluating against fragment manually annotated neutrality; these indicate accuracy about 20%