Our approach for knowledge graph question answering GRASP (https://github.com/ad-freiburg/grasp) internally uses search indices built from knowledge graphs to work. Ideally, the search functionality provided by these indices should directly be available to GRASP via the used knowledge graph engine QLever itself.

This project is about:

• Adding embedding support to QLever to enable similarity search
• Adding prefix keyword search support to QLever to enable prefix-keyword search