Sentiment Analysis and Emojification of Tweets

Abstract

– Social media platforms have become a prevalent means for individuals to share their emotions and thoughts. With millions of tweets being posted on Twitter every day, these tweets provide us with a vast dataset. Conducting sentiment analysis on this dataset can be a valuable method to obtain meaningful insights about societal trends. For this purpose, a sentiment analysis model and a web interface that emojifies emotions were developed using the Python programming language. This model works on tweets shared on Twitter and utilizes natural language processing techniques to determine the sentiment of the tweets. In this study, 168.274 English tweets were collected using the Twitter API. The collected tweets underwent a cleaning process where URLs, hashtags, mentions, and emojis were removed. Then, the TextBlob Python library was employed to label the tweets as positive, negative, or neutral. The labeled tweets were subjected to classification accuracy testing using Gradient Boosting, Logistic Regression, Naive Bayes, Random Forest, and Support Vector Machines machine learning models. The findings revealed that logistic regression achieved the highest classification accuracy with 94%. Lastly, a web interface was developed, which retrieves the last 50 tweets of a queried user's profile and appends a relevant emoji based on the sentiment of each tweet.