Sentiment analysis and classification by considering negation polarity shifter and opinion summarization for product reviews

Sentiment analysis is special kind of task in Natural language processing as it determines the sentiment expressed by the customer or different users in the form of text format. There are different techniques existed to determines the user sentiment into different polarities. BOW it is very popular model used to represent and model the text for sentiment analysis and text classification. It gives better performance for text classification but sometimes BOW model performance becomes worst in sentiment classification, because it has some limitations and disadvantages. The main challenging task in BOW model is polarity shift problem. We proposed a system to address the polarity shift problem to provide feasible solution to the BOW model and make use of BOW model in sentiment analysis with better performance. In this system we are addressing, removing and modifying polarity shift problem in context of sentence-level sentiment analysis and phrase-level sentiment analysis in the contrast of negative reviews. Furthermore we extend our system to determine sentimental features or opinions that has expressed by the customers in given comments (in text format) and opinion targets. At last we performed classification on customer reviews to classify that reviews into three sentiment polarities i.e., positive, negative, neutral class labels and generate graph based on opinion summarization for customer understanding purpose. Proposed system based on lexicon-based approach using dictionary based technique.