A Hybrid Framework for Scalable Opinion Mining in Social Media: Detecting Polarities and Attitude Targets

Text mining of massive Social Media postings presents interesting challenges for NLP applications due to sparse interpretation contexts, grammatical and orthographical variability as well as its very fragmentary nature. No single methodological approach can be expected to work across such diverse typologies as twitter micro-blogging, customer reviews, carefully edited blogs, etc. In this paper we present a modular and scalable framework to Social Media Opinion Mining that combines stochastic and symbolic techniques to structure a semantic space to exploit and interpret efficiently. We describe the use of this framework for the discovery and clustering of opinion targets and topics in user-generated comments for the Telecom and Automotive domains.