knowledge-extraction

name	knowledge-extraction
description	Directory index for knowledge-extraction: knowledge-extraction
version	1.0.0
license	MIT
author	Synthos
metadata	{"synthos":{"signature":"paper_content: str, schema: dict -> structured_knowledge: dict (entities, relations, claims, evidence)","atom_type":"skill","priority":"P1","related_skills":[]}}

Knowledge Extraction

详细内容请加载此skill后按需执行。核心流程和命令已提炼如下：

快速参考

详细文档和完整命令列表已被移至 references/ 目录以保持简洁。 This skill has been compressed. Full content is available in references/.

IO_CONTRACT

• input: paper_text: str, query_focus: str
• output: extracted: list[KnowledgeExtract] — 包含 topic, method, dataset, metric, result, confidence, source, quote

对应原则：P2（机械原子暴露输入输出规范）

---

Knowledge Entry Generation Mode

This skill also governs the knowledge_entry step of the research pipeline (step 4/4: literature_scan → gap_analysis → hypothesis_generation → knowledge_entry). When loaded for knowledge_entry generation, produce a synthesis document that integrates the outputs of all prior steps.

IO_CONTRACT (knowledge_entry mode)

• input: candidate_id: str, hypothesis_generation_output: dict, gap_analysis_output: dict, literature_scan_output: dict — outputs from all 3 prior pipeline steps
• output: knowledge_entry: dict — markdown file with 9 required sections, 6-dimension quality score

Required Sections

A knowledge_entry markdown file MUST contain the following sections in order:

#	Section	Content	Source
1	Pipeline Completion Summary	Table with step scores (literature_scan, gap_analysis, hypothesis_generation, knowledge_entry) + score evolution (HG→KE delta + rationale)	All prior steps
2	6-Dimension Quality Scoring	Weighted scoring table with 6 dimensions (see reference knowledge-entry-rubric.md). Each dimension needs score + rationale paragraph. Show weighted total calculation.	Novel synthesis
3	Domain Overview	Natural history of the physiological system (3-5 paragraphs), clinical problem table (condition, prevalence, existing metric, gap), white space statement, "why now" justification	literature_scan + gap_analysis
4	Model Architecture	2-ODE+PINN description: ODE-1 and ODE-2 dynamics (state vars, parameters, key behavior). ASCII diagram or equation blocks. Multi-scale temporal assessment table. Cross-ODE identifiability confound analysis (detection→assessment→mitigation→residual).	gap_analysis
5	Entity Extraction	Two tables: (a) Entities — name, type, description for all physiological systems, diseases, parameters, devices, kernels; (b) Relations — source, relation, target for structural_analog/neighbor, clinical_biomarker, integrates, etc.	All prior steps
6	Hypothesis Portfolio	Ranked table (rank, ID, title, composite, recommendation) + detailed description of primary hypothesis (statement, population, measurement, endpoint, rejection criteria, rationale) + discriminative experiment design (Pattern #5 phase table)	hypothesis_generation
7	Kernel Registration	Structured table: kernel ID, name, type, domain, state/parameters, dynamics, clinical inputs, connected kernels, unique contribution, clinical conditions	Novel synthesis
8	Clinical Translation & Market	Phase 1-4 pathway table (timeline, activity, success criterion). Addressable populations table (population, size, current test, PINN advantage). Zero-equipment justification.	gap_analysis + hypothesis_generation
9	Critical Assessment & Limitations	Star-rating table (novelty, data feasibility, clinical impact, model complexity, parameter identifiability, reproducibility) + key risks (3-5 items, each with mitigation). Pipeline completion summary (domain expansion trajectory, next actions).	All prior steps

Score Evolution Pattern

The knowledge_entry's 6-dimension weighted score is systematically 1-4% below the hypothesis_generation step score. This is EXPECTED and normal — the 6D rubric weights conservative dimensions (Methodological Soundness 0.20, Reproducibility 0.10) that don't appear in the 5-dimension hypothesis scoring rubric.

State of gap_analysis	Typical Δ (HG → KE)	Notes
PASS (≥0.85)	−1 to −3 points	6D rubric restores conservative judgment
CONDITIONAL (0.65-0.78)	−2 to −4 points	Larger delta because methodological concerns get more weight in 6D

6-Dimension Quality Scoring Rubric

See references/knowledge-entry-rubric.md for the full rubric with detailed scoring criteria per dimension per score band.

Examples

Existing knowledge_entry files follow this exact format. Reference examples:

• outputs/papers/_knowledge_only/cochlear-mechanics-PINN/knowledge_entry_cochlear-mechanics-PINN.md (auditory, K-013)
• outputs/papers/_knowledge_only/baroreflex-regulation-PINN/knowledge_entry_baroreflex-regulation-PINN.md (cardiovascular, K-010)
• outputs/papers/_knowledge_only/respiratory-sinus-arrhythmia-PINN/knowledge_entry_respiratory-sinus-arrhythmia-PINN.md (cardiopulmonary, K-011)
• outputs/papers/_knowledge_only/vocal-fold-phonation-PINN/knowledge_entry_vocal-fold-phonation-PINN.md (laryngeal, K-014 — most recent example)