Twitter act as a most important medium of communication and information sharing. As tweets do not provide sufficient word occurrences i.e. of 140 characters limits, classification methods that use traditional approaches like “Bag-Of-Words” have limitations. The proposed system used an intuitive approach to determine the class labels with the set of features. The System can able to classify incoming tweets mainly into three generic categories: News, Movies and Sports. Since these categories are diverse and cover most of the topics that people usually tweet about .Experimental results using the proposed technique outperform the existing models in terms of accuracy.

In this paper, a novel idea is being presented to perform Opinion Mining in a very simple and efficient manner with the help of the One-Level-Tree (OLT) based approach. To recognize opinions specific for features in customer reviews having a variety of features commingled with diverse emotions. Unlike some previous ventures entirely using one-time structured or filtered data but this is solely based on unstructured data obtained in real-time from Twitter. The hybrid approach utilizes the associations defined in Dependency Parsing Grammar and fully employs Double Propagation to extract new features and related new opinions within the review. The Dictionary based approach is used to expand the Opinion Lexicon. Within the dependency parsing relations a new relation is being proposed to more effectively catch the associations between opinions and features. The three new methods are being proposed, termed as Double Positive Double Negative (DPDN), Catch-Phrase Method (CPM) & Negation Check (NC), for performing criteria specific evaluations. The OLT approach conveniently displays the relationship between the features and their opinions in an elementary fashion in the form of a graph. The proposed system achieves splendid accuracy across all domains and also performs better than the state-of-the-art systems.