Ranking System for Opinion Mining of Features from Review Documents
Tanvir Ahmad, M. N. Doja · 2012
With the exponential growth of the web there has been explosive increase in the user generated contents in the form of customer reviews, blogs, discussion forums, social networks etc. Most of the contents are stored in the form of unstructured or semi structured data from where distillation of knowledge is a challenging task. In this paper we propose a feature wise opinion mining system which first extracts features from user generated contents, then determines the intensity of the opinions by giving emphasis to the modifier of the words, which expresses opinions. It finds the numeric score of all the features using Senti-WordNet and then calculates the overall orientation of the feature to determine how intense the opinion is for both the positive and negative features. The positive and negative features are identified by extracting the associated modifiers and opinions. The summary is presented by specifying the features in descending order of importance.