MISSION · TechnicalMachine learning
ES
LEVEL SELECTED
EmotionSense
Emotion-aware responses for conversational AI.▼ SCROLL TO BEGINMAIN OBJECTIVE
Standard chatbots often respond without accounting for the emotional context behind a user’s words.
Detect emotional context and examine how emotion-aware responses differ from standard outputs.
Used the GoEmotions taxonomy and reviewed classification errors to understand ambiguous and overlapping labels.
Compared standard conversational behavior with an emotion-aware layer as a product differentiation opportunity.
Balanced emotion-label detail against prediction reliability and a clear user-facing experience.
WHAT I BUILT
Built a Python pipeline using RoBERTa, Hugging Face, PyTorch, scikit-learn, and Streamlit to classify emotion and compare responses.
Produced an interactive technical prototype connecting model behavior to a human-centered product use case.
Model accuracy alone is insufficient; uncertainty, tone, and response behavior determine whether the product feels trustworthy.
PythonRoBERTaPyTorchHugging FaceStreamlit