Language Lineage

The problem

Most readers meet a language family as a settled fact. West Germanic simply counts as separate from Proto-Germanic, and nobody explains why some scholars still reject "Altaic" as a family at all. That context exists in academic sources: Glottolog, sound-law catalogs, individual monographs. But it's written for other linguists, scattered across specialist literature no curious reader can just follow.

I built Language Lineage for people with no linguistics background: anyone curious about where their language's ancestors came from. Historical-linguistics classification doesn't resolve the way an open math problem might; more research won't settle whether "Altaic" is a family. So the site shows live disagreement, not a single tidied-up tree: who argues what, and how settled each claim is.

What I built

A branch can descend from more than one ancestor

Contact-driven change, such as creoles, sprachbunds and areal diffusion, can give a branch more than one ancestor. I modeled that as a DAG from the first commit: each branch stores a list of parent ids, and the layout algorithm was built from day one to handle more than one parent, not patched in later.

Every disputed classification carries both sides' evidence

Macro-family disputes such as Altaic, Nostratic and Austric are live scholarly disagreements, not solved-but-unpublished problems, so picking one side would misrepresent the field. Each disputed grouping is its own record: a status (widely accepted, minority position, largely rejected), the evidence for it, and the evidence against it, drawn on the tree with a dashed border and a ring-count glyph.

I hand-wrote the branch layout instead of reaching for a graph library

Branches can have more than one parent, so a plain linear timeline wasn't enough, but each family's tree stays small, up to a dozen branches, which didn't justify a general graph-layout dependency. The layout computes each branch's layer, then orders and refines it to keep the tree legible, drawn as the growth rings I wanted. I came back and adjusted it more than once as family trees grew wider than the original pilot.

Content lives as JSON files in the repo, with no database or backend

Every family and branch is one JSON file, so a correction is a single-file pull request. It's the same model I use across my other static reference sites, no admin panel, no login. The whole site builds to static HTML with no server to run and no accounts to secure.

The build itself catches data errors before they ship

With 165 hand-authored branch files, a bad id or a citation issue would otherwise only surface as a broken page after it deployed. The validator rejects an unknown parent, a cycle in the branch graph, and, added later, any turning point dated after every source cited for it. A broken commit can't reach production.

Result

Language Lineage is live. It maps 18 language families from proto-language to present, with 602 sourced turning points, 77 disputed classifications each showing both sides' evidence, and 67 historical figures indexed by the events they're tied to. I built it solo over five days in September 2026, writing every family's content myself, then wiring the world map and sound-law comparison views on top of the same data.