Emotion Detection Based on Text and Emojis

Opinion mining has become increasingly vital in today's digital world for making strategic decisions. With the volume of information increasing daily at a fast pace it becomes necessary to refine the information to efficiently analyze important and very vital data. Analysis of text to extract the emotions that are conveyed is a very important task, as analysis of text manually is time-consuming as well as may contain errors. Emojis have played a significant role in the digital world for communication and conveying various opinions. To solve these challenges an emotion detection model is introduced, that aims to address the issues related to accuracy of recognizing emotions by combining textual data and emojis. This model aids in various fields like sentiment analysis, mental health monitoring, feedback assessment. This research begins its journey by the creation of a diverse data set that contains text-based content that are meticulously labeled. To prepare the content in the data various preprocessing tasks are undertaken such as tokenization, removal of stopwords, normalization of text. The model is built in a recurrent neural network that is designed for managing multimodal input and trained through multi-label classification to predict the appropriate emotion category. Empirical formulas are referred to validate the model's effectiveness, which help in comparing the results obtained from text-only models in emotion recognition and emojis related sentiments. The aspects that set this model apart from the other models is its ability to adapt across platforms and languages, by recognizing the changing digital communication nature. As the nature of digital communication keeps on evolving it is necessary that various analytical tools are also evolved in parallel to address the changes as they come.