A Survey on Analysis of Sentiment Using Twitter Dataset

In this modernizing era, the elevation of web technology and its proliferation have built a web network on which lies a vast volume of data for the users of internet, and a lot of data is being formed. The Internet has progressed as a modernized framework for exchanging ideas, skill development, and communicating opinions through online mediums. Social networking sites like Instagram, Twitter, and Facebook have gained an immense amount of popularity due to their features that allow people to interact, express their points of view, share beliefs about certain topics, have discussions, or post messages, images, and videos across the internet, connecting the globe. There has been enormous work done in the area of Twitter data analysis for sentiment. This research work focuses mainly on emotion classification of Twitter data, which helps in analyzing the information shared through tweets in which opinions can be extremely heterogeneous, unstructured, and also can be positive or negative, or neutral in some scenarios. In this study, we will supply the survey and a performance comparative analysis of present processes for analyzing sentiment, like Machine Learning, Deep Learning, and lexicon-based methods, by evaluating their performance metric, i.e., accuracy. Evaluation of techniques will include Machine Learning algorithms like Support Vector Machine, Naive Bayes, Decision Tree, Random Forest, and, along with the deep learning algorithms which include Recurrent Neural Networks, Long Short-Term Memory Networks, and Hybrid Approaches, and many more. After evaluation of past techniques, we will provide the best technique based on the accuracy of existing opinion mining models, comprising traditional methods and recent enhancements in AI.