Mood to Action Recommender
This project implements a hybrid, AI-driven, content-based recommendation system that uses natural language processing to infer a user’s emotional state and intent from free-text input.
Github Link!
Overview
This project uses machine-learning models that are used for emotion and intent detection, while a rule-guided recommendation engine safely maps these signals to relevant actions. Semantic sentence embeddings are then applied to rank recommendations, producing explainable and context-aware suggestions focused on wellbeing and productivity.
Technologies
• Python
• Pandas
• Scikit-learn
• Sentence-Transformers (MiniLM)
• Cosine Similarity
• Custom NLP Models for emotion and intent detection
Features
• Emotion Detection using NLP models trained on labelled emotional text
• Intent Classification to identify user goals (e.g. seeking advice, planning, reflection)
• Hybrid Recommendation Engine with rule-guided filtering and content-based logic
• Semantic Ranking using transformer-based sentence embeddings
• Explainable Recommendations with human-readable reasons
