Source Intake
Candidate APIs and datasets are tested in small reproducible batches.
Scientific infrastructure for cannabinoid evidence
MaryGenAI
Project for scientific data sources, ontology design, and human-reviewed curation of cannabinoid therapy evidence.
Source evaluation
Ontology design
Human review
Evidence Record
Field-level provenance snapshot
- Publication Type
- Systematic Review
- Study Design
- Human Clinical Trial
- Compound
- Cannabidiol (CBD)
- Review Status
- Human verified
Provenance
Each extracted claim is linked to a source record, extraction method, confidence score, evidence text, and reviewer state.
The Knowledge Graph
A structured ontology for cannabinoid evidence.
MaryGenAI models publications, clinical trial records, compounds, receptors, ligands, medical conditions, routes of administration, dosing objectives, adverse events, and human review state as connected evidence objects.
Publication
PMID / DOI / abstract
Compound
CBD / THC / CBG
Study Design
RCT / cohort / in vitro
Receptor
CB1 / CB2 / TRPV1
Dosing Objective
Safety / effective dose
Curation Pipeline
Designed for provenance before scale.
Every extracted field is treated as a claim with source, method, confidence, evidence text, and review status. Human review is tracked at field level, not only at record level.
Normalization
Records are mapped into publication, trial, interaction, and evidence claim objects.
Assisted Extraction
Models classify fields only when evidence and confidence can be stored.
Human Review
Reviewers verify field values, corrections, notes, timestamps, and ontology version.
Latest Research
Semantic Scholar API mockup.
This interface sketch shows how MaryGenAI could distinguish publication type from study design while preserving source identifiers and ontology-ready fields.