🔤 Onoma: Every Word, Traced
Dictionaries tell you what a word means. Onoma lets you investigate it. It is a client-side workbench for English vocabulary that treats a word the way a professional does — through its senses, the concepts it belongs to, its frequency across genres, its pronunciation in thirteen dialects, and where it came from. It is English-only, account-free, and, after the first load, mostly offline.
What it does
- Five ways to search. Find a word by spelling, by what it means, by a letter pattern, by rhyme, or by "sounds like".
- A full word page. Senses with provenance, IPA, etymology, frequency rank with a genre breakdown, register tags, domains, and a dialect-by-dialect pronunciation strip.
- Synsets and their relations. Synonyms, antonyms, hyponyms, meronyms and hypernyms, with a glossary that explains the relationship you are looking at.
- A workbench. Compare saved words side by side and run set operations — intersection, union, difference — over their relation closures, then export the result as TSV.
- A library. Saved words and saved searches, persisted locally in IndexedDB.
The corpus is the hard part
The app is the easy half. The data underneath it is assembled by a dependency-free Python pipeline from pinned sources — Open English WordNet 2025, Wiktionary via Kaikki, the COCA frequency list, and ipa-dict — and normalised into gzipped JSON chunks that the browser inflates on demand with DecompressionStream. Senses, IPA and etymologies carry a per-record source, so provenance is a property of the data rather than a claim in a footer. The pipeline's own verify.py refuses to pass on truncated chunks, entries filed under the wrong letter, or lost synonymy — checks that previously let hollow records ship.
The app is deliberately honest about its edges: some entries are structural shells, the meaning index cannot yet reach Wiktionary-only senses, and the "sounds like" feature is substring-and-rhyme rather than phonetic. Those are written down, measured, and queued, not hidden.
Tech stack
| Technology | Role |
|---|---|
| Ionic React + React | App shell, navigation and UI |
| TypeScript | Strict application code |
| Vite | Build and dev server |
| Python | Reproducible corpus pipeline |
| vite-plugin-pwa | Installable, offline-capable app shell |
The result
Onoma is in pre-production, with the corpus and the client both complete and passing CI — lint, types, tests, a shuffled re-run, the build, and the pipeline's integrity checks. It is the studio's deepest bet on the studio's own thesis: owning a dataset of real linguistic value and shipping it to the browser with nothing between the reader and the word.