Encoder-decoder based multi-label emoji prediction for Code-Mixed Language (Hindi+English)

Abstract

Emojis enjoy an important place in digital communication. They can express feelings and emotions in contexts when words cannot. In other words, they add emotions to a piece of text. Emojis are rising concurrently with the increased use of social media platforms for communication and have become a language in itself. Single emoji prediction systems are no longer adequate because multiple emojis are being grouped to convey emotions these days. The multi-label emoji prediction system for code-mixed language has not yet been explored to the best of our knowledge. It explores multi-label emoji prediction in Hinglish, one of the most commonly used code-mixed languages. This paper presents a framework for Hinglish multi-label emoji prediction. The proposed Encoder-decoder based Emoji Prediction model for Hinglish (EDEPHi) model outperforms other baseline models and is far more diverse in terms of predicted emojis.