Review on Sentiment Analysis Techniques in Data Mining Domain: Exploring Sentiment Analysis Techniques and Research Gaps in Data Mining

Abstract

We know that the internet is a collection of networks, and the use of the internet has changed the way people express their thoughts and feelings. People are connecting with each other with the help of the internet through blog posts, online conversation forums, and many more. Sentiment analysis is mainly concerned with the identification and classification of opinions or emotions of each post. Sentiment analysis is broadly classified in the two types first one is a feature or aspect based sentiment analysis and the other is objectivity based sentiment analysis. To correctly classify the tweets, machine learning technique uses the training data. So, this technique does not require the database of words like used in knowledge-based approach and therefore, machine learning technique is better and faster. Several methods are used to extract the feature from the source text. Feature extraction is done in two phases In the first phase extraction of data related to twitter is done i.e. twitters specific data is extracted. Now by doing this, the tweet is transformed into normal text. In the next phase, more features are extracted and added to feature vector. Each tweet in the training data is associated with class label. This training data is passed to different classifiers and classifiers are trained. Then test tweets are given to the model and classification is done with the help of these trained classifiers. So finally, we get the tweets which are classified into the positive, negative and neutral. In this paper, our goal is to present the study on different sentiment analysis methods and feature extraction methods designed by various researchers. Additionally, we are presenting the current research gap based on analysis of all recent methods of sentiment analysis. The outcome of this paper is the current research problems and motivation for sentiment analysis in data mining.