littleseneca
ad187c9857
Reduced orphans from 311 (65.3%) to 246 (51.7%), edge density from 0.41 to 0.87. Connected 65 previously orphaned claims. Clusters processed: - grade+treatment: 110 edges from 6 batches (103 target claims) - cd-treatment/uc-grade: 13 edges from 2 batches - cross-domain (microbiome/nutrition/diagnosis/pathophysiology): 70 edges from 5 batches Key improvements: - Pathophysiology domain: 0 → 67 edges (now 6.7 edges/claim, fully connected) - Contradicts edges: 2 → 10 (surfaced SIBO diagnostic debates) - Depends_on edges: 5 → 19 (guideline lineage documented) - Extends edges: 46 → 189 (cross-guideline treatment recommendations connected) Scripts: phase3b_generate_candidates.py, phase3b_apply_batch.py, phase3b_crossdomain.py Edge batches: 11 edge JSON files in scripts/manifests/phase3b_candidates/ Verified: 0 broken wikilinks, 0 schema violations, 0 duplicate pairs
OSKG-IBD
An open-source knowledge graph for Inflammatory Bowel Disease (IBD) and Small Intestinal Bacterial Overgrowth (SIBO), built on the OSKG methodology. Structured evidence synthesis connecting pathophysiology, diagnosis, treatment, microbiome science, clinical guidelines, and dietary interventions into a typed-claim graph.
What This Is
A structured decomposition of the IBD/SIBO literature into discrete, verifiable claims with typed edges (supports, contradicts, extends, depends_on). Think of it as a machine-readable evidence map — every claim traces to its source, every edge documents the relationship between claims.
Why IBD and SIBO
IBD affects millions globally. SIBO is increasingly recognized as a key driver of IBS symptoms and a potential co-morbidity in IBD patients with persistent symptoms despite remission. The literature spans gastroenterology, immunology, microbiology, nutrition science, and clinical guidelines — domains that rarely talk to each other. A knowledge graph bridges them.
SIBO is particularly interesting as a knowledge graph subject: its definition is contested, its diagnosis methods are debated, and its relationship to IBD is an active area of research. The graph captures this uncertainty explicitly via confidence ratings and contradictory edges.
Pipeline
| Phase | Artifact |
|---|---|
| 0 | Source acquisition and extraction |
| 1 | Chapter/section reading notes |
| 2 | Claims extraction with intra-batch edges |
| 3 | Cross-source edge construction |
| 4 | Synthesis: hinge inventory, cascade trees, structural gaps |
| 5 | Capstone: evidence-grounded synthesis |
Repo Structure
notes/
├── pathophysiology/ Disease mechanisms, genetics, immune dysfunction
├── diagnosis/ Endoscopy, biomarkers, breath testing
├── treatment/ Pharmacological, biologic, surgical
├── microbiome/ Gut flora, dysbiosis, SIBO/IMO mechanisms
├── nutrition/ Diet interventions, EEN, SCD, LFE
├── clinical-guidelines/ ACG, AGA, ECCO, Rome Foundation
├── history/ Evolution of IBD/SIBO understanding
├── claims/ Discrete claim nodes
├── evidence-briefs/ Synthesized evidence on specific questions
├── questions/ Open research questions
└── synthesis/ Capstone documents
sources/
├── books/ Textbooks and patient guides
├── guidelines/ Clinical practice guidelines
└── papers/ Research papers and reviews
See BOOK-GUIDE.md for the full source corpus.
License
This project is for research and educational purposes. Source texts remain under their original copyrights.