Collaborations Workshop 2026 of de-RSE
Searching semantic data management systems and knowledge graphs is a challenging endeavour. Prevalent semantic query languages like SPARQL are difficult to learn and therefore pose a high barrier to widespread adoption among non-technical users.
Large language models (LLMs) provide a promising way to enable researchers without advanced skills in SPARQL or other domain-specific search languages to perform specific data queries that can be used for subsequent analysis and automation.
We have developed an LLM-based query interface for the open-source semantic data management system LinkAhead. Its main component is a translator that converts natural language into LinkAhead’s domain-specific query language, CQL. The component can be used with a variety of open-source, open-weight, or proprietary LLMs.
Hallucination risks are mitigated through a combination of constrained query generation and human-in-the-loop verification. A graphical interface enables users to inspect generated queries in a structured form before execution, reducing the likelihood of incorrect queries being sent to the LinkAhead backend.
This workshop will introduce the design and architecture of the LLM-based translator. Furthermore, we will present validation results and discuss limitations.
With: Dr Alexander Schlemmer