Natural Language Processing (NLP), MLOps, Azure
Project Overview
The Emotion Detective project aimed to create an emotion-detection model to analyze Expeditie Robinson episodes, helping Banijay Benelux better understand viewer engagement. Using a custom Python package, the project classified emotions like happiness and sadness, aiming to uncover audience response patterns. This automated model processes sentence-level emotions, offering targeted insights.
Pipeline and Deployment
The model was deployed on Azure, using a custom pipeline for ingestion, training, and inference. By processing raw videos through NLP models like RoBERTa and RNN, the setup ensured scalable, automated emotion analysis across large datasets.
Skills and Challenges
Throughout development, key skills in NLP and cloud deployment were strengthened, with challenges primarily in managing cloud infrastructure and refining audio quality. The final tool is accessible to Banijay Benelux, providing enhanced emotion-driven insights for content creation.
The package can be installed by using pip install emotion_detective
Click here to access the package
Click here to access the documentation
Acknowledgement
This project was a collaborative effort with my team members: