Sentiment Mining through Mixed Graph of Terms

摘要: The spread of social networks allows sharing opinions on different aspects life and daily millions messages appear the web. This textual information can be divided in facts opinions. Opinions reflect people's sentiments about products, personalities events. Therefore this is a rich source data for opinion mining sentiment analysis: computational study opinions, emotions expressed text. Its main aim identification agreement or disagreement statements that deal with positive negative feelings comments reviews. In paper, we investigate adoption probabilistic approach based Latent Dirichlet Allocation (LDA) as Sentiment grabber. By approach, set documents belonging to same knowledge domain, graph, Mixed Graph Terms, automatically extracted. paper shows how graph contains weighted word pairs, which are discriminative classification. proposed method has been tested standard datasets real-time analysis tweets holders various contexts. experimental evaluation effective satisfactory.