Noisy Review Mining

Abstract

Opinion mining is the task to identify the user opinion about a particular object. So, the opinion mining tool processes the reviews collected from different reviewers by generating a list of object/product features and performing aggregation of the opinions about each feature. With the advent of Web 2.0 user got freedom to interact with other users and can create their own web contents. The use of social networking sites (such as Facebook, Twitter, Orkut) and SMS for text messaging has increased a lot recently, which lead to emergence of new language aimed for compactness and time saving. As a result, thousands of abbreviations and shortcuts are used regularly. Now a days, Social networking sites and online shopping sites like Amazon.com are used by the users to express their opinion about products, events, people etc. As the reviews available on these sites may contain noise such as misspellings, typos, non-standard abbreviations, stylish writing, there is need to make the data noise free so that it can be used for opinion mining. In this work a framework has been proposed to perform opinion analysis of noisy reviews using techniques such as Term similarity and Document frequency. The reviews for different products have been tested by this framework and the corresponding result is displayed in Negative (-ve) and Positive (+ve) form. The results are found to be satisfactory for all the tested products.