Enigmatrix studies regulatory non-compliance as a connected information problem: changes are discovered late, procedures are difficult to understand, risk remains invisible, and unsupported claims circulate without traceable evidence. Four research modules share a versioned evidence layer that moves an official rule from publication to understandable, targeted SME action.
One evidence chain
V6 fixed-split test
Information-aware model
Common 200-post test
The programme is structured around evidence contracts between modules—not four disconnected applications.
Awareness gap
Regulatory change awareness
My research module
Knowledge gap
Compliance knowledge accuracy
Experimentally evaluated
Risk invisibility
Compliance risk visibility
Research evidence
Misinformation
Regulatory claim verification
Experimentally evaluated
Each module owns a distinct barrier, dataset and evaluation question. Module 1 is my individual contribution.
Awareness gap
Regulatory change awareness
Can regulatory changes be discovered, extracted, classified and delivered early enough to measure and reduce SME awareness lag?
Gazette corpus, page-aware extraction, classification, grounded trilingual summaries, alerts and lag measurement.
Knowledge gap
Compliance knowledge accuracy
Can verified regulatory evidence become accurate, complete and grounded procedural guidance?
Knowledge instrument, verified procedure corpus and retrieval-grounded guidance research.
Risk invisibility
Compliance risk visibility
Which business and information-access variables explain future compliance failure?
SME vulnerability dataset, interpretable prediction and explanatory information factors.
Misinformation
Regulatory claim verification
How prevalent is regulatory misinformation, and which evidence-grounded verification approach works best?
Annotated claim corpus, spread analysis and evidence-grounded claim-checking research.
Deterministic services carry high-liability work; model and GPU branches remain isolated behind explicit evidence gates.
Read the full research and implementation record: questions, dataset lineage, multilingual extraction, model selection, experimental RA-HMT architecture, grounded summaries, deployment boundaries, negative results, and open field gates.
Open the Module 1 dossier