STRUCTSENSE: A Task-Agnostic Agentic Framework for Structured Information Extraction with Human-In-The-Loop Evaluation and Benchmarking
Researchers have developed STRUCT-SENSE, an open-source framework designed to improve structured information extraction from scientific literature. This task-agnostic system combines ontology-guided symbolic knowledge with agentic self-refinement and human-in-the-loop validation. Evaluations across tasks like schema-based extraction, metadata extraction from papers, and neuroscience NER demonstrated its generalization capabilities and accuracy, even extracting additional entities beyond gold annotations in some biomedical benchmarks. AI
IMPACT This framework could accelerate scientific discovery by improving the extraction of structured data from research papers.