Tag
The paper proposes a retrieval-augmented generation framework using large language models to recommend safety countermeasures from crash narratives, achieving high precision and recall in evaluation on real-world data.