Yiddish XT9
A compact, offline-first Yiddish text-entry tool: a numeric keypad mapped to Yiddish letters, with a local prediction engine that turns digit sequences into ranked Yiddish words and composes them right-to-left — packaged for Android with Cordova.
Mobile engineer — Cordova app
Where it started.
T9 — typing words by tapping a numeric keypad — was built for Latin alphabets: each digit holds a few letters and software predicts the word. None of that maps to Yiddish. The letters group differently, the text runs right-to-left, and the prediction has to know Yiddish words, not English ones.
The need was specific and unglamorous: let someone type Yiddish quickly from a numeric keypad — hardware or on-screen — and have it predict the right word. No cloud, no account; it had to work on the device, offline.
Off-the-shelf keyboards don't cover this. It needed a purpose-built keypad mapped to Yiddish letter groups, a right-to-left composer, and a prediction engine that understood Yiddish and ran entirely on the phone.
How I thought about it.
I built it as a single-screen Cordova Android app: a custom 12-key keypad mapped to Yiddish letter groups, a right-to-left output area, and a browser-side prediction engine. Key presses become a T9 digit sequence, matched against a Yiddish word list bundled with the app.
The prediction pipeline is the heart of it. An exact digit-sequence match returns the candidate words, ranked by how common they are; when the exact sequence isn't enough, Levenshtein distance finds nearby sequences and folds those in as suggestions; and pressing 0 cycles through the alternatives. The whole 15,626-word dictionary ships in the app bundle, so it all happens offline — no server, no latency.
The thing itself.
- Yiddish T9 keypadA 12-button keypad with Yiddish letter groups, plus delete, space, and symbol handling — built for numeric-keypad input.
- Right-to-left composerThe output area is explicitly RTL and updates word-by-word as you type.
- Predictive T9 engineMaps each digit sequence to dictionary entries and ranks exact matches by popularity.
- Fuzzy correctionLevenshtein distance suggests nearby T9 sequences when the exact one isn't enough.
- Candidate cyclingPressing 0 cycles through the alternative predictions for the current sequence.
- Offline dictionaryShips a 15,626-entry Yiddish word list with popularity and T9 codes in the app bundle — no network.
What changed.
Yiddish words appear as you tap the keypad, predicted from a local dictionary and composed right-to-left.
It runs fully offline — the dictionary and the prediction pipeline live in the app bundle, so there's no server, account, or latency.
Generic numeric input is replaced with Yiddish-specific letter mapping, ranked prediction, and candidate cycling.
It's packaged for Android with Cordova — a small, focused utility rather than a heavy app.
Outcomes are qualitative — no client metrics shown.
Under the hood.
Got a process that
should run itself?
Tell me where the work piles up. I'll show you what a calmer, quieter version of it looks like — built around how your team already works.