Overview
An end-to-end architecture encompassing trip detection, on-device behaviour analysis, and a dual-component NLG pipeline designed for robustness to intermittent connectivity and noisy sensor data. The system uses a two-step "reflection" pipeline for higher-quality, psychologically and culturally appropriate feedback delivered after trips. Pilot deployment with 90 drivers demonstrated promising evaluation results.
Technology
- Python
- NLG
- Machine Learning
- Android
Themes
- AI
- Natural Language Generation
- Nigeria
- Driving Behaviour