Incident sheet · /enterprise-ai

Filed
Read
3 min · 593 words
Severity
Routine

India Unifies 65 Heritage Portals Behind One AI Search Platform

India's Ministry of Culture has unified 65 heritage portals behind one AI semantic search platform, using RAG and Bhashini to answer in 22 languages, with citations and no login.

Ministry of Culture launches AI platform to unify 65 digital heritage portals
FeedMinistry of Culture launches AI platform to unify 65 digital heritage portals — sasint / pixabay

Event log

  1. Ministry of Culture unified 65 digital heritage portals into one AI-powered semantic search platform launched by Secretary Vivek Agrawal
  2. The platform uses embeddings, a knowledge graph and Retrieval-Augmented Generation, and is LLM-agnostic on an open-source foundation
  3. It supports all 22 scheduled Indian languages via Bhashini, with voice input and spoken output
  4. It queries only official ministry portals — no open internet — and provides numbered citations; live queries cover Gyanbharat, Abhilekh Patal, Indian Culture and Mera Gaon Meri Dharohar
  5. No login or registration is required for public access via culture.gov.in

India's Ministry of Culture has connected 65 separate digital heritage portals into a single AI-powered semantic search platform, launched by Ministry Secretary Vivek Agrawal and live now through the ministry's main portal, culture.gov.in.

Until now, the ministry's holdings — archives, manuscripts, museums, monuments, art forms, preservation projects and national campaigns — sat scattered across dozens of independent systems. A researcher looking for Ayurveda manuscripts, for instance, had to know which portal held them and which keywords that portal's search box expected. The new platform removes that guessing game: one conversational query runs against all 65 portals at once.

The takeaway up front: a fragmented public records estate has become a single queryable knowledge base, at no cost to the user and with no account required.

How does the search actually work?

The platform is not a keyword engine with a chat window bolted on. It combines three techniques from current enterprise AI practice:

  • Embeddings — converting text into numerical representations so the system can match meaning, not just strings of characters.
  • A knowledge graph — structured links between people, places, documents and artefacts across the 65 portals.
  • Retrieval-Augmented Generation (RAG) — fetching verified source material first, then generating an answer grounded in it.

Natural Language Processing parses the question. The system returns contextually relevant results enriched with AI-generated summaries, numbered citations, multilingual translations and multimedia content. Each citation links directly to the ministry's official pages, so users can verify every claim against the source record.

A user can ask, in plain language: "Which manuscripts on Ayurveda are held in national collections?" The platform collates relevant material from each portal into a single coherent answer rather than a list of links. It also understands follow-up questions in context, so a user can drill into a topic step by step without restating earlier information.

The engineering choice here matters for longevity: the platform is LLM-agnostic and built on an open-source foundation, which lets the ministry swap in emerging Indian sovereign AI models as they mature rather than locking itself to one vendor's model.

Where does the data come from?

The platform queries only the ministry's official portals. It does not search the open internet. When the underlying data is thin, the system says so instead of filling gaps with generated text — a deliberate guardrail against hallucinated answers in a government records context.

For the ministry's largest national databases — Gyanbharat, Abhilekh Patal, Indian Culture and Mera Gaon Meri Dharohar — the system runs live queries at the moment of search, which keeps even deep archival records reachable. Relevant images, artefacts and documents appear alongside the text of each answer.

Users choose between brief and detailed answers depending on their needs.

What about language and access?

The platform supports all 22 scheduled Indian languages, incorporating the national language translation initiative. It is built on Bhashini, the Government of India's national language platform.

It accepts voice input and speaks answers back in natural Indian voices — a design choice aimed at users with low literacy or visual impairments, for whom typed queries and dense text are barriers.

Access requires no login and no registration. No personal profile is created. Anyone can query the national cultural record immediately and anonymously.

For CIOs, the deployment is a working template: an open-source, model-agnostic RAG architecture, citation-backed answers drawn solely from authoritative internal sources, and live federation across dozens of legacy databases — the standard checklist for trustworthy enterprise knowledge retrieval.

Sixty-five portals, twenty-two languages, one question box.

via thenewsmill.com (Source)

Filed under

  • retrieval-augmented-generation-rag
  • enterprise-search
  • knowledge-graphs
  • digital-government
  • multilingual-ai
Share this article:

More from Amara Osei

Amara Osei

Show full bio

Correspondent covering regulation, policy and their impact on enterprise information technology and the leaders who run at CIO Monitor.

188 articles

The newsroom digest

One email a day with the stories that matter. No noise.

By subscribing you agree to the privacy policy.

Cross-references

‹ Previous articleNext article ›