Modelling emoji use by Maltese speakers on Twitter

Abstract

This dissertation studies the semantics of emojis as used by Maltese users on Twitter. Using Tweepy and Snscrape in Python, four datasets of Maltese English tweets were gathered. For each dataset, a Word2Vec model was created to perform the analysis and create word embeddings. Further analysis was performed by asking human subjects to associate English words with the 10 most frequent emojis in the dataset. In general, the word embeddings performed relatively well and some interesting results were observed. The results by the assessors were similar to some degree. Furthermore, a t-SNE projection was created to visualize the emojis. This results in four different clusters of emojis.