From Emoji Usage to Categorical Emoji Prediction

Emoji usage drastically increased recently, they are becoming some of the most common ways to convey emotions and sentiments in social messaging applications. Several research works automatically recommend emojis, so users do not have go through a library thousands emojis. In order improve emoji recommendation, we present distribute two useful resources: an embedding model from real usage, clustering based on these embeddings identify groups Assuming that emojis part written natural language can be considered as words, only used unsupervised learning methods extract patterns knowledge tweets. Thereby, emotion categories face were obtained directly text fully reproductible way. These resources methodology multiple usages; for example, could our understanding or enhance recommendation.