Hebrish
Experimental Hebrish is an early-stage proof of concept, trained on a small code-switching dataset in one speaker's voice. First results are promising. Growing the training data is a long-term effort, and the aim is for every version to get better. Roadmap

Send words

Send a word list

Noticed Hebrew words that dictation keeps getting wrong, or that are missing from the lexicon? Send them as a list. Every list is read by a person and considered for the next round of training sentences.

How to format the list

  • Comma-separated: arnona, kupat holim, makolet. One word or phrase between each comma.
  • Alphabetical, with no duplicates.
  • Latin letters, spelled the way you would write it in an English message. If you know the Hebrew spelling or the meaning, put it in the description or the note, not in the list.

You don't need to do this by hand: paste the words in any shape (one per line, separated by commas or semicolons, a bulleted list) and press Tidy the list. It separates them with commas, removes duplicates and sorts them A–Z. You can then edit the result before you send it.

One word, a bug, or something else? Use the contact form.