| Evolution of Web 2.0 and social media has enabled people to express their thoughts and opinions thorough various platforms such as blogs, discussion forums, tweets and other apps. Apps like Yelp and Zomato collect users' data and furnish the reviews on a particular business to help new users to make decisions. Performing sentiment analysis on such social media user data to extract their opinions and feelings on various things has been quite fruitful. This project aims to develop new social media sentiment analysis algorithm for businesses competition. We apply topical text classification on users' data from Yelp to create a model for competition analysis in the interest of business operators. Topical text classification is different from sentiment based classification since it involves collecting data on a specific item/activity which is more useful based on the user interest. The proposed algorithm promotes multi-label classification by mapping the Yelp review corpus into 4 categories ("Food", "Service", "Deals" and "Ambience") and performs sentiment analysis based on Tree-LSTMs, to extract item specific opinions. This project evaluates the performance of the proposed algorithm in terms of accuracy, precision, and validity. |
