A Neurosymbolic Approach for Constructing Planning Domain Models from Clinical Narratives
NSPIN turns unstructured surgical notes into probabilistic planning models for clinical workflow support.
The paper says the system uses a pretrained LLM to extract and fill in event sequences from raw laparoscopic appendectomy narratives. It then induces a PPDDL model and uses LLM-proposed precondition revisions checked against empirical validation. The authors tested it on 2,660 notes from 9 surgeons, reporting generalization to unseen notes. Clinical expert review found the induced knowledge largely aligned with surgical practice. ArXiv · AI/CL/LG's note
The paper says the system uses a pretrained LLM to extract and fill in event sequences from raw laparoscopic appendectomy narratives. It then induces a PPDDL model and uses LLM-proposed precondition revisions checked against empirical validation. The authors tested it on 2,660 notes from 9 surgeons, reporting generalization to unseen notes. Clinical expert review found the induced knowledge largely aligned with surgical practice. ArXiv · AI/CL/LG's note
score 4