Abstract:
Problem: With the increasing concern around mental well-being, there is a growing need for
systems that can understand user emotions in real time and provide comforting responses.
This project introduces MindBeats, an emotion-aware music recommendation system that
leverages a chatbot to interact with users, detect their emotions through text-based
conversations, and offer personalized Spotify music recommendations to uplift their mood.
Methodology: The system utilizes a hybrid deep learning model combining Convolutional
Neural Networks (CNN) and Bidirectional GRU layers for emotion detection from user text
inputs. The architecture is built using Python (Flask) for the backend and React for the
frontend. Text preprocessing includes tokenization and padding. Follow-up questions are
dynamically generated using the Gemini API based on detected emotions, enhancing the
interaction. The final mood prediction is computed using emotion trends across all user
responses, followed by curated music playlist recommendations.
Initial Results: The implemented emotion classification model achieved an accuracy of
96.09%, with precision, recall, and F1 scores all around 0.96, indicating strong predictive
performance and reliable emotion recognition.
