The Future of Feelings: Leveraging Bi-LSTM, BERT with Attention, Palm V2 & Gemini Pro for Advanced Text-Based Emotion Detection
Harsha Vardhan Khurdula, Anbilparithi Pagutharivu, Jin Soung Yoo · 2024
This research explores advanced text based emotion detection by leveraging Bidirectional Long Short-Term Memory (Bi-LSTM), Bidirectional Encoder Representations from Trans-formers (BERT) with attention mechanisms, and the latest in Generative Artificial Intelligence, including Palm V2 and Gemini Pro. Unlike traditional supervised learning approaches that rely on pre-defined labels, our methodology employs generative models to adaptively recognize a broad spectrum of human emotions in real-time. By finetuning generative AI, we aim to surpass current approaches in emotion detection accuracy and provide a more nuanced understanding of textual sentiment. Our work outlines the novel approach, dataset preparation, model architecture adjustments, and comprehensive evaluation metrics to demonstrate the efficacy of combining these technologies for enhanced emotion detection.