Evidence-backed research demo
Knowledge Prism helps scholars, research centres, think tanks, and policy teams turn scattered documents into claim reviews, knowledge maps, and audit-ready research briefs.
Research archives sprawl across PDFs, notes, citations, and folders.
Scholars, labs, think tanks, supervisors, and policy research teams.
Knowledge maps, review queues, sample briefs, and audit trails.
Claims stay tied to source relevance, evidence level, and review status.
Loading current project facts...
Problem
Researchers have PDFs, Zotero libraries, policy reports, notes, and grey literature scattered across folders. Search can surface documents, but it cannot tell what is trustworthy, what is relevant, what is still only a clue, what needs expert review, or what can enter an explanatory graph.
Who it is for
Built for PhD scholars, supervisors, research centres, think tanks, library heads, and policy research teams who need more than a document dump. They need a map of claims, evidence, uncertainty, and interpretive value.
Demonstration snapshot
Method
The scholar keeps the question and the final judgment. The Human Workbench turns source clues into guarded outputs without silently promoting them into verified ontology.
Inventory the corpus and expose what is only metadata.
Sample text before treating a generated claim as usable.
Keep each claim tied to its current level of support.
Ask what the text does for the research question.
Connect works, concepts, interactions, and cautions.
Carry the review state forward with provenance intact.
Use cases
Turn scattered reading into a defensible map of arguments, sources, and open questions.
Know what your archive contains, what is usable, and what remains uncertain.
Keep policy claims tied to evidence, source, and review status.
Separate relevant signals from background noise across reports and briefs.
Move from catalog records to interpretive, auditable knowledge infrastructure.
Outputs
Trust layer
Every claim keeps its source, evidence grade, relevance decision, functional interpretation, and review status. Metadata can suggest a path, but it is not treated as scholarship. The graph grows only where the evidence can follow.
Trustworthy by design
The Human Workbench accepts a research question, source clue, evidence grade, and functional role, then exports a review note while keeping ontology promotion blocked until provenance is strong.
Visitors begin with the value, then move into the pilot demonstration, method, evidence, and expert machinery only when needed.
The Expert Dashboard remains available for deep inspection, but it is clearly marked as the advanced evidence cockpit.
The public demo uses safe exports only. Raw retrieval outputs, raw corpus files, local paths, secrets, and database files are not served.
Machinery, when needed