llmsherpa can infer parent/child relationships from nlm-ingestor outputs. Incorporate llmsherpa logic / code to our parser pipeline to create appropriate relationships between layout annotation blocks. This will allow us to do advanced retrieval techniques like finding appropriate text and then crawling up the annotation tree to the parent section and feeding that to the LLM.
llmsherpa can infer parent/child relationships from nlm-ingestor outputs. Incorporate llmsherpa logic / code to our parser pipeline to create appropriate relationships between layout annotation blocks. This will allow us to do advanced retrieval techniques like finding appropriate text and then crawling up the annotation tree to the parent section and feeding that to the LLM.