How TamizhConnect Works: Fragments → Family Tree

Most family-tree tools want dates, birth certificates, and cleanly-typed names. That works fine for families whose records survived intact. Tamil families across South Africa, Malaysia, Fiji, Mauritius, the Gulf, Singapore, Sri Lanka, the UK and Canada usually don't have any of that. They have a grandfather's initials, a village name their mother once argued about, and a story about a ship.

TamizhConnect is built to start there. Below is what actually happens after you sign up.

What most people already know

Most new users start with three or four fragments. "Appa's initials were R. K." "We're from somewhere near Thanjavur." "Periappa left for Malaysia around Independence." Sometimes a black-and-white photo with a name written in Tamil on the back, half-faded. That's the whole inheritance.

You don't need to tidy any of it up before you begin. Type it in as-is. If you know the year within a decade, that's enough. If a village name has three spellings your relatives argue over, add all three. The matcher deals with fragments — sanitising them into false precision is what breaks it.

How the name-matcher reads Tamil

Tamil names don't fit the Western surname-plus-forename shape, so the matcher goes at it differently. Something like "R. K. Meenakshi" gets stored as-is; matching compares the given name and the last-token surname across candidate records, and cross-checks the father-name field on other records against those tokens. Colonial-era passport spellings — Meenaachy, Meenatchi, Meenakchi — get pulled in through a phonetic variant table that expands the common Tamil sound-swaps (doubled consonants, aa/a variants, sh/s, dh/d).

That's usually enough to catch the case where one record spells a name one way and another spells it another. It's not deep patronymic parsing though. The matcher won't infer that a later "K. Kumar" is the son of a "Krishnan" from the K prefix alone. If you know that relationship, add it directly.

Nothing here rewrites your data. The matcher surfaces suggestions — "this person on your tree looks like this person in the indenture archive" — and you accept, ignore, or defer.

The datasets we check against

Several hundred thousand arrival records are indexed across the diaspora archives we've digitised or partnered with. The main ones:

• theiii.in (III) — the largest source, covering arrivals to South Africa, Fiji and Suriname primarily. Nested JSONB records with per-record destination fields.
• South Africa Natal (GLDC) — records from the Gandhi-Luthuli Documentation Centre at UKZN, covering Natal arrivals from 1860 onward. The Truro arrived in Durban on 16 November 1860 after a roughly 10-12 week voyage from Madras.
• Fiji (Girmitiya Madras) — records tracing the Madras Presidency emigration to Fiji (indenture ended there in 1916).
• UK Passenger Lists (TNA) — outward voyage records from The National Archives at Kew (BT 26 / BT 27 series), useful for tracing 20th-century onward migration.

Live per-provider counts are on the indenture search landing page. They change as new archives get ingested.

On the Indian side, we index Tamil Nadu electoral rolls at the village level. That's mostly there to help a diaspora family check whether "the Kumaresan family from Ramanathapuram district" is still traceable, and to work out which modern village name matches the historical one a grandfather remembered.

Some communities aren't covered by the arrival archives yet — Trinidad, Guyana, Kenya, Réunion, and a few others. Trees for those families still work; you just won't get automatic hints from those particular archives until we ingest them.

Where hints stop and you take over

Hints aren't conclusions. Every suggestion is labelled with the record it came from, and you can open that record to check it against what you know. This matters because indenture records are messy — a plantation clerk in 1892 wrote down what he heard, a British administrator re-transliterated it a decade later, someone digitised it in the 1990s under a fourth spelling. Accept a hint without checking and you'll end up with a tree that quietly fuses two different people into one, or has a father dying before his son was born.

Rough rule of thumb: accept when the village, decade, and family structure all line up. Ignore when only the name matches (that usually means it's a common name, not the same person). Leave it pending when the village matches but the generation doesn't quite, and come back to it after talking to an elder.

Nothing bad happens if a hint sits pending for months. Some of ours do. That's fine.

The tree and the migration map

Once you've added people and started resolving hints, the platform builds two views from what's there.

The tree view is what most people expect — parents, children, siblings, spouses, with a lineage-focus mode that follows the direct ancestors and descendants of the home person when the whole thing gets too wide to read.

The migration map is a separate view, drawn from the origin and destination places you've entered on people. It's gated on paid plans (the migration_map_personalized entitlement) — on the free tier you'll see a sample map rather than one built from your own tree. For families on a plan that includes it, the map draws the route your family took: village of origin, intermediate stops, and where each branch ended up. It's the view that tends to get screenshotted and shared on WhatsApp with elders — it makes migration legible to younger relatives who glaze over when you talk about ancestry.

Sharing with relatives

The tree is single-owner today. Trees don't have a shareable-link feature yet, and there's no mode where an invited relative can view or edit your tree from their own account. If a cousin wants to see what you've built or wants to contribute a photo or a memory, the practical options are:

• Screenshot the tree view and send it. Works for one-off conversations.
• Publish a Family Story — a public narrative built from specific people on the tree — and share that link. You control which people appear and what details are visible.
• Export the tree as GEDCOM and share the file if the relative already uses a genealogy tool.
• Add the cousin's contribution yourself to your tree, crediting them in a note.

Multi-user editing and per-person comment threads aren't shipped. If that changes we'll update this section.

What's private, and what isn't

Every tree starts private. Nobody sees it unless you share a link.

Living people are treated more strictly than deceased ancestors — anyone without a death date has their details hidden by default from external search and shared views. One thing to know: publishing someone into a Family Story (a public narrative you choose to put on the platform) exposes the specific details you include in that story, whether or not the person is living. Only publish what you'd be comfortable being read by strangers.

Trees are encrypted in transit over HTTPS. We don't run third-party trackers on authenticated pages, don't sell contact details, don't run ads inside trees, and don't offer "premium reveal" access to other people's families. Revenue comes from subscriptions, not from selling data about the families using the platform.

What families end up with

Every tree grows at its own pace. Some families build most of it in a weekend and add a few names a year afterward; others treat it as an ongoing project, adding a person when a photo turns up or a cousin mentions someone at a wedding.

The ones that get the most out of it share a habit: they don't try to finish. They add what they know, mark what they don't, and come back when a new fragment surfaces — a photo cleared out of a shoebox, a name mentioned at a wedding, a passport found while sorting through papers after a funeral. Because the tree lives with you long-term, small additions compound into real coverage.

Next steps

If you want to start:

• How to trace your Tamil roots walks through a first-visit workflow with real fragments.
• Find Tamil ancestors by village covers the village-lookup side in more detail.
• The Four Anchors method is the mental model we use when nothing lines up yet — village, kula deivam, community, migration route.

If you already have a tree started on Ancestry, MyHeritage, or FamilySearch, GEDCOM import isn't built yet — it's on the Q2 2026 roadmap. For now, add the top 3-5 people manually to get started and expand from there. GEDCOM export from TamizhConnect does work, so you can pull data out to another tool if you need to.