Information Extraction
Documentsinformation-extractionInformation Extraction is an Agent Skill for documents published in the beita6969/ScienceClaw repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Extract structured entities, relations, and clauses from scientific and legal documents
QUICK START
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/beita6969/ScienceClaw/blob/HEAD/skills/information-extraction/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/information-extraction, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Information Extraction
Purpose
Extract structured information (entities, relations, events, clauses) from unstructured scientific and domain-specific text.
Key Datasets
- ChemProt (bigbio/chemprot): Chemical-protein interaction extraction from BioCreative VI; 10 relation types (CPR:3-CPR:9) between chemicals and proteins
- CUAD (atticus-project/cuad): Contract Understanding Atticus Dataset; 41 clause types from 510 legal contracts (CC-BY licensed)
- JNLPBA: Biomedical named entity recognition (protein, DNA, RNA, cell line, cell type)
- SciERC: Scientific entity and relation extraction from AI paper abstracts
Protocol
- Schema definition — Define target entity types, relation types, and attributes
- Preprocessing — Sentence segmentation, tokenization, abbreviation expansion
- Entity recognition — Identify and classify named entities (NER)
- Relation extraction — Detect relationships between entity pairs (RE)
- Normalization — Map entities to standard ontologies (MeSH, ChEBI, UniProt)
- Output structuring — Format as structured JSON, RDF triples, or knowledge graph
Extraction Types
- Chemical-protein interactions: Substrate, inhibitor, agonist, antagonist, activator
- Legal clause extraction: Termination, IP rights, non-compete, indemnification, limitation of liability
- Gene-disease associations: Causal, biomarker, therapeutic target
- Drug-drug interactions: Synergistic, antagonistic, pharmacokinetic
Rules
- Report extraction confidence scores for each entity/relation
- Provide span offsets for traceability back to source text
- Normalize entities to standard identifiers (CAS, UniProt ID, etc.)
- Handle nested entities and overlapping relations
- Validate extracted facts against known databases when possible