Etymology, made visual

Words change shape and meaning as they move through time and languages. Drawn as a tree, word origins are easier to follow, and easier to remember.

Built from English Wiktionary, CC BY-SA. More than 10 million entries from over 5,100 languages.

One word, its whole etymology

Search calme, the French word for calm, and its ancestors are laid out as a tree: Middle French, Italian, Latin, back to Ancient Greek καῦμα, "burning heat". Each node is coloured by language. Each line carries two things from the source data. The label says what kind of link it is: inherited, borrowed, derived, an influence. The line style says how certain that link is: solid when it is certain, dashed when probable, dotted when uncertain.

The example word, calme, traces back through:

  1. calme (French)
  2. calme (Middle French)
  3. calma (Italian)
  4. cauma (Latin)
  5. καῦμα (Ancient Greek)
  6. καίω (Ancient Greek)
  7. *káyyō (Proto-Hellenic)

How it's built: The tree is assembled from Wiktionary's etymology records, followed link by link through the database, and drawn from those records. Because the source texts are written in prose, a language model is used to read them and turn them into nodes and links. The tree is meant to keep the sources' nuances: a link they call probable or uncertain is drawn that way, not as a certainty.

Click any word in the tree

Each node opens on its own entry: the original Wiktionary etymology text, its senses, its relatives. The Descendants tab turns the tree around: from Ancient Greek καῦμα, you can see the words recorded as growing out of it, across languages, and follow any of them to its own tree.

How it's built: Descendants are read from the database by following the same parent-to-child links in the other direction. The list is as complete as Wiktionary's records are.

A short history, with its sources

A few sentences tell the story in order: what changed, where, and roughly when. Every claim carries a numbered reference to the Wiktionary entry it comes from. Where part of a word's history is uncertain, debated, or simply unknown, the open questions below the text say so.

How it's built: The history is condensed from the Wiktionary etymology texts of the words in the tree, and nothing else. A language model reshapes those texts into a few readable sentences, and the numbered references point back to the texts so you can check each one.

The same idea across language families

Translations are grouped by family, so patterns show up at a glance: calma across the Romance languages, spokój and its cousins in Slavic, śānti through the Indo-Aryan ones. Each one opens its own tree. For English words the translations come from Wiktionary; for other languages they are generated and marked as such.

Keep the words you like

Sign in with Google to save words to collections: for a language you are learning, a topic, or simply the ones that stayed with you. A collection can be shared with a link.

How it's built

  1. Gather. Each search starts in a database built from Wiktionary. The word's ancestor chain is followed link by link, and the etymology text of every word on it is collected and cleaned of asides.
  2. Reshape. Those texts are prose. A language model changes their form. It produces the nodes and links of the tree, a few sentences of history, and a list of the points the sources leave uncertain or debated. The system never relies on the language model's knowledge of etymology.
  3. Draw and check. The nodes and links are then assembled into a single tree and checked as a whole before it appears on the page. Results are saved, so the next visitor gets them instantly.

Etymology data from Wiktionary, CC BY-SA 4.0 · About Miraris