Extracting ontology-compliant knowledge from scientific text describing irradiated materials using large language models
The paper introduces eolas, an LLM pipeline that turns irradiated-materials papers into ontology-aligned knowledge graphs.
The authors say the system targets data buried in unstructured scientific text, especially for materials meant to withstand fusion-reactor heat and radiation. They report that eolas can produce high-quality graphs in minutes, compared with an expert taking roughly 30 to 90 minutes per article. The output is shown in tables with faceted navigation so humans can validate it. The paper also presents a benchmark dataset and analyzes 168 experiments across models and prompting methods. ArXiv · AI/CL/LG's note
The authors say the system targets data buried in unstructured scientific text, especially for materials meant to withstand fusion-reactor heat and radiation. They report that eolas can produce high-quality graphs in minutes, compared with an expert taking roughly 30 to 90 minutes per article. The output is shown in tables with faceted navigation so humans can validate it. The paper also presents a benchmark dataset and analyzes 168 experiments across models and prompting methods. ArXiv · AI/CL/LG's note
score 4