Sentiment Analysis and Sarcasm Detection of Indian General Election Tweets

TL;DR

This paper presents a transfer learning-based approach to analyze sentiment and detect sarcasm in Indian election tweets, addressing the challenge of sarcasm which was previously overlooked.

Contribution

It introduces a novel method combining transfer learning and SVM with TF-IDF for sentiment and sarcasm detection in Indian election tweets.

Findings

Effective sentiment analysis of election tweets

Successful sarcasm detection in social media data

Enhanced model performance with transfer learning

Abstract

Social Media usage has increased to an all-time high level in today's digital world. The majority of the population uses social media tools (like Twitter, Facebook, YouTube, etc.) to share their thoughts and experiences with the community. Analysing the sentiments and opinions of the common public is very important for both the government and the business people. This is the reason behind the activeness of many media agencies during the election time for performing various kinds of opinion polls. In this paper, we have worked towards analysing the sentiments of the people of India during the Lok Sabha election of 2019 using the Twitter data of that duration. We have built an automatic tweet analyser using the Transfer Learning technique to handle the unsupervised nature of this problem. We have used the Linear Support Vector Classifiers method in our Machine Learning model, also, the Term Frequency Inverse Document Frequency (TF-IDF) methodology for handling the textual data of tweets. Further, we have increased the capability of the model to address the sarcastic tweets posted by some of the users, which has not been yet considered by the researchers in this domain.