Uniting the Power of Text and Emojis to Revolutionizing Sentiment Analysis of Twitter Data

This research innovates sentiment analysis by incorporating emoji data alongside text collected from Twitter, which points to the problem of text-based analysis in the identification of discrete emotional contexts. A new multimodal sentiment analysis model was created using LSTM networks based on a dataset of 199708 tweets that can be positive, neutral or negative. The proposed approach also improved the classification of sentiments to an overall accuracy of 90.20%, with positive sentiments and neutral sentiments being well classified, although there was difficulty in classifying negative sentiments. Due to the presence of textual as well as emoji features it can generate more profound emotional interpretation and can potentially be used in social media analysis and survey analysis. This research is useful to the field because it shows that combining multiple types of sentiment analysis is can enhance the comprehension of emotion in computer-mediated interaction.