200,000 Government Resolutions. One unified pipeline.
Whether scanned on paper, trapped in legacy font encodings, or natively digital — all of Maharashtra's GRs currently sit inaccessible.
The challenge
Three eras of documents. One pipeline handles them all.
Pre-2010: Scanned paper
The bulk of the historical archive exists only as scanned images. Full OCR is required, and generic tools struggle with Marathi.
2005-2015: Legacy font trap
PDFs look like Devanagari but use proprietary ASCII mappings — Shree Dev, Kruti Dev, Shusha. Text extraction produces garbled output.
Post-2015: Digital but inaccessible
Natively typed in Unicode, text is extractable — but no GR is searchable, indexed, or cross-referenced.
Mixed encoding everywhere
Single PDFs contain Unicode headers, legacy-font bodies, and image-based tables. No clean cutoff exists.
See it in action
Documents in. Verified text out.
2007
Pre-2010 — Scanned / legacy encoded, 2 pages
The outcome
Every GR, connected and searchable.
Entity extraction turns every GR into a node in a knowledge graph. Departments, officials, laws, budgets, vendors, and prior resolutions are all linked — making Maharashtra's institutional memory queryable for the first time, by government, auditors, journalists, researchers, and the public.
A single GR may touch a dozen entities. Across 200,000 resolutions, that's millions of connections — queryable by anyone.
In practice
What this actually makes possible.
Not another archive to browse. Specific, everyday questions that were effectively impossible to answer before — each one now a single query. Trace a budget to its last vendor. Surface a policy conflict before a new GR ships. Find every decision that touches a village.
Find every decision affecting your area
Search by district, taluka, or village. See every GR that mentions your region — with dates, departments, and decisions in one place.
Track how a policy has evolved
Follow a subject — teacher recruitment, crop insurance, reservation — from its first GR through every amendment and reversal.
Follow public money end-to-end
Trace a budget allocation from the GR that announced it to every vendor, contractor, and project it funded.
Find precedent in seconds
Before drafting a new GR, see every related prior decision — who approved, what they cited, what the outcome was.
Spot cross-department conflicts
Check whether a proposed GR overlaps or contradicts what other departments have already decided. Catch conflicts at draft time.
Preserve institutional memory
Every administration's decisions stay connected to every other. Continuity doesn't depend on any individual's memory.
And much more. We're only scratching the surface of what becomes possible once every decision is connected.