A novel dictionary-based classification algorithm for opinion mining

There has been a rapid rise in the number of users getting connected online via social networking sites. To communicate with other users and share their thoughts and opinions, online users' tend to use texts in the form of blogs, posts, tweets, messages, reviews, comments etc. Thus, there has been an immense possibility complemented with a wide gamut of research in the field of Opinion Mining or Sentiment Analysis by using textual information from online communities. Hence, there is an extensive need for different text classification algorithms and approaches to classify texts and predict sentiments correctly so as to comprehend the emotional state of the user. We have varied algorithms for text classification for predicting emotional traits. In this paper, we are proposing a novel dictionary-based algorithm that uses lexicon-based approach for opinion mining and calculates the sentiment polarity levels. Our algorithm is different from other lexicon-based algorithms in the context that it uses the three degrees of comparisons viz. positive, comparative and superlative degrees on words; for each of the positive and negative sentiment words. Our system yields an Accuracy of 81% and an F-score of 0.874 on the test dataset which is quite moderate and can be fairly accepted.