Twitter Based Emotion Recognition Using Bi-LSTM

Twitter and other social media platforms have evolved into crucial sources of information and communication for billions of people across the world. The massive amount of data collected on these platforms enables insights into people's emotions and feelings. Emotion identification, in particular, has gotten a lot of attention because of its potential uses in industries like marketing, healthcare, and social study. "Emotional states of individuals, also known as moods, are central to expressing thoughts, ideas, and opinions, which influence attitudes and behavior." In this study, we present a method for identifying the emotion or mood of a tweet and classifying it into the appropriate emotional category. The work mentioned in this paper aims to identify and analyze the mood and emotion displayed by people in their text tweets to offer ideas. We gathered tweets and replies on a particular set of themes to create a dataset with text, user, emotion, sentiment, and other data. We used the dataset to compute user impact ratings based on various user and tweet-related factors to analyze feelings and emotions in tweets and their replies. Our model used a brand-new bidirectional long short-term memory (BiLSTM) and previously learned word embedding vectors. Finally, using the latest data, we made generalized and specific consumer suggestions based on Twitter's conduct. According to the experimental findings, the suggested model has 95.56% and 99.16% prediction accuracy. Modern performance improvement was achieved for the model.