Rule Based Sentiment Analysis System

Abstract

Due to the exponential enhancement in the Internet usage and replacement of public opinions, sentiment analysis becomes an important process in today’s life. Sentiment analysis is a process of extracting information from user’s opinions. Every person shares his or her information in social network sites, blogs, product review websites and webforums. Thus, the thoughts of other people provide information that helps in decision making process. But, sentiment analysis is a challenging task because it is very difficult to find the exact sentiment from text as there are so many challenges like entity identification, subjectivity detection in performing sentiment analysis. The project work that has been carried out in the thesis is focused on implementation of rule based sentiment analysis system. Earlier, only adjectives were used as feature to perform sentiment analysis. But, this system uses adverb-adjective combinations as a feature as it improves the sentiment analysis process. The system extracts the twitter posts and computes the frequency of each word in tweet. Then, it calculates the sentiment and score of each tweet. The system also computes the sentiment and score of simple English sentences. Some of the challenges like thwarted expectations, pragmatics, and entity identification have been resolved by system. The system has also used UNL (Universal Networking Language) to resolve the challenges like entity identification etc. At the end, the results of simple English sentences given by the system have been compared with manual testing of these sentences.