International development projects are usually tagged to a specific sector, such as “Health” or “Water and Sanitation”. However, for real-world planning and coordination, much more detail is needed. Over the past 5 years, AidData.org has worked to do detailed “activity coding” on development projects – taking them from “Health” to “Malaria control” or “Medical Insurance Programs” – based on written project descriptions.
The challenge is now for developers with skills in natural language processing, machine learning, and/or algorithm design to automate some of this coding. A special github.io page will be created to compare the performance of different submissions.
Outcomes: tool to provide automatic activity coding based on project descriptions.
Main datasets: reference data sets with defined answers.
Lead: David Rust-Smith, AidData