LFS automation: Generating ground truth data
This project supports the development of an AI-based system for automatically classifying Labour Force Survey (LFS) responses into standard occupational and industry codes. To benchmark and improve the performance of the large language model (LLM), a high-quality ''ground truth'' dataset of human-coded responses will be generated.
The project recruits and trains Zambian graduates in economics, statistics, and related disciplines to serve as annotators: by combining human annotation with automated evaluation, the project will deliver a robust training and validation dataset.
