Using sentiment analysis to improve emoji predictions in tweets

Fani Noncheva


Supervised by Luis Espinosa-Anke; Moderated by Matthias Treder

This project is focused around using sentiment analysis to improve the task of emoji prediction in tweets. Given a string of text, the machine learning model must learn to find patterns that would yield different predictions as to what emoji would follow. Generally, people use different emojis to express similar emotions, so to improve emoji predictions in tweets would be to observe the differences in the way people express themselves via text and how it affects the emojis they use. To keep the sample size manageable I intend to only focus on the twenty most used emojis. The goal of the project is to find if there is a correlation between specific types of sentiment and differing use of similar emojis. In order to achieve the goal, I intend to use and evaluate different machine learning algorithms and a variety of sentiment analysis tools in python.

Initial Plan (31/01/2020) [Zip Archive]

Final Report (05/06/2020) [Zip Archive]

Publication Form