International Journal of Engineering and Technology

Social networking sites prove to be indispensible tools for decision making owing to the large repository of user views accumulated over a period of time. Such a real data can be exploited for various purposes such as making buying decisions, analysing the user views about new product launched by a company, product promotion campaign , impact of policy decisions made by a political party on society etc. In the current work the authors have proposed a generic model for feature based polarity determination by sentiment analysis of tweets. This model has been implemented by the seamless integration of R tool, XML, JAVA, Link Parser A practical multistep system, in place, efficiently extracts data from tweet text, pre-process the raw data to remove noise, and tags their polarity. Data used in the current study is derived from online product feature based reviews collected from tweeter tweets. Link parser version 4.1 b is employed for parsing a natural sentence which is broken into multiple tokens corresponding to noun and adjective before being stored in a persistent storage medium. The objectivity score is determined using SentiWordNet 3.0 lexical resource which is parsed using a tool implemented in Java. The linguistic hedges are taken care of using Zadeh’s proposition which modifies the final objectivity score. The objectivity score so computed, provides the necessary guidelines in influencing decisions. The authors have tested the model for product purchase decisions of two different sets of products, smart phone and laptop based on predefined set of features. The model is generic and can be applied to any set of products evaluated on a predefined set of features.