WSSA: Weakly Supervised Semantic-based approach for Sentiment Analysis

Wissam Mammar Kouadri, Salima Benbernou, Mourad Ouziri, Iheb Ben Amor

Abstract: In this work, we propose a Weakly Semantic-based approach for Sentiment Analysis (WSSA), a novel approach that analyzes sentiment by considering weak labels from different sources (sentiment analysis tools) and aggregates them based on features such as the consistency between sources, the semantic equivalence between documents, and experts’ domain knowledge in order to improve the sentiments analysis tools results. The aggregation is achieved using a Probabilistic Soft Logic reasoner to infer the documents’ polarity.