Today I’m putting something small and free out into the world: Wordhoard, a morpheme dictionary.

It’s a searchable catalog of 13,659 English morphemes: the prefixes, suffixes, combining forms, and roots that words are actually built from, each with its senses, plain-language definitions, real example words, etymology, and a note on where it’s attested. You can search by form (pyr-), by meaning ("fire"), or by an example word, filter by affix type or etymological family, and follow a morpheme to its other senses and its cousins. The whole dataset is free to download (CSV and JSONL) under CC BY-SA 4.0.

It lives at word-hoard.lexid.online (with the data hosted at morpheme-dictionary.lexid.online).

Why a morpheme dictionary?

Wordhoard grew out of Lexid, a game where you cast spells by building words: fusing prefixes, roots, and suffixes into “wirds” with real grammatical logic. To make that game honest, I needed a clean, trustworthy map of English morphology: what each piece means, what it attaches to, where it comes from, and whether it’s a real attested thing rather than something I made up.

That map turned out to be useful well beyond the game. A word list tells you that “incomprehensibility” exists. A morpheme list tells you why: in- + com- + prehend + -ible + -ity, negation wrapped around grasping, made abstract. So I cleaned it up, stripped out the game-design bits, rewrote anything that needed rewriting into original wording, gated every row on real attestation, and published the lexicography layer on its own.

The long chain

Here’s the thing I keep coming back to: I didn’t really make this. I assembled it, and only because an enormous, mostly invisible crowd of people, across a very long time, made the pieces I assembled from.

Lexicography is one of the great relay races of human effort. It runs from Pāṇini cataloguing Sanskrit morphology twenty-five centuries ago, through the Greek and Latin grammarians, through Johnson and Webster, through the thousands of anonymous volunteer readers who mailed slips of paper to the Oxford English Dictionary for decades. Every generation hands the baton on a little more complete than it received it. Nobody finishes the work; everybody adds a link.

This dataset is one more small link, and it only exists because of the links right before it:

• Wiktionary and Wikipedia: the staggering volunteer effort behind the definitions, etymologies, and example words, released under CC BY-SA so people like me can build on them.
• Kaikki.org / wiktextract (Tatu Ylönen): for turning Wiktionary’s sprawling markup into clean, machine-readable data, which is the only reason a one-person project could touch a corpus this size.
• Open English WordNet and the Princeton WordNet: George Miller, Christiane Fellbaum, and the lexical-semantics tradition that taught all of us to think about meaning as a network of senses.
• Wikidata Lexemes: released into the public domain (CC0), which rescued a long tail of rare classical roots that nothing else attested.

And, more diffusely: every etymologist, morphologist, and dictionary editor whose careful, unglamorous work became the ground truth all of these sources rest on. Thank you. Genuinely. This is yours more than it is mine.

A necessary caveat

I should be plain about one thing: I’m not a trained linguist. I’m a designer and lifelong word-nerd who fell down the morphology rabbit hole, and everything here leans heavily on the published scholarship and open datasets named above. They are the authority; my job was mostly to compile, reconcile, and rewrite their work carefully. I checked what I could, but in a corpus this size, assembled by a single non-specialist, errors of judgment, oversimplifications, and plain mistakes are inevitable, and where they exist they are mine, not my sources’. So if you’re leaning on Wordhoard for anything that matters (research, teaching, publication), please treat it as a well-built starting point and verify against the primary sources, rather than as an authority in its own right.

Pay it forward

Because I stood on open data, Wordhoard is open data too. The full dataset is CC BY-SA 4.0: download it, fork it, fix my mistakes, fold it into your own project. If you do, the deal is simple and the same one I was offered: keep the attribution, link the license, note your changes, and pass the next person a corpus a little better than the one you got. That’s the whole chain, working as designed.

The license (CC BY-SA 4.0) already asks for attribution; this is just how to give it. If Wordhoard helps your work, and especially if that work is scholarly or published, the one courtesy I ask in exchange for the free use is a citation, to me and to the dataset:

Park, Logan O. Wordhoard: An Open Morpheme Dictionary. 2026, https://word-hoard.lexid.online. Licensed under CC BY-SA 4.0.

Or, in BibTeX:

@misc{park2026wordhoard,
  author = {Park, Logan O.},
  title  = {Wordhoard: An Open Morpheme Dictionary},
  year   = {2026},
  url    = {https://word-hoard.lexid.online},
  note   = {Open lexicography dataset. CC BY-SA 4.0}
}

It costs nothing, and it means a great deal to a one-person project; attribution is exactly what keeps small, free, open resources alive long enough to be worth building on.

It’s incomplete, of course it is. There are senses to split, etymologies to deepen, roots still waiting to be attested. That’s not a bug; that’s the job description. If you spot something wrong, I’d love to hear it.

For now: go look up a word’s bones. Try tele-, or -monger, or gn (as in know, cognition, diagnosis: same ancient root, hiding in plain sight). The fun of morphology is that once you start seeing the pieces, you can’t stop.

— Logan