Turtles all the way down

My word processing application automatically corrects some typos. For example, if I write “newtork” it knows I really meant to write “network” and it makes the correction for me. For people like me who make lots of typos, this is a great feature.

But if the word is not in the spell checker’s dictionary, the word processor can’t seem to figure out how to correct it. I recently had to write the word “estoppel” a bunch of times in a document. The application should have, after seeing the word several times: (1) either given me an opportunity to put the word in the dictionary or automatically done so, and (2) add it to the words that will now be automatically corrected.

Also the same lines, if the application observes me making the same typo over and over, and it doesn’t know how to automatically correct this typo, it should figure out how to do so from context. That is, it should learn from my own corrections of this word, then add my corrections to the automatic correction database.

This would also work for names with uncommon spellings and words in other languages.

There has been work on similar concepts in the spell checking of search engine requests. I’ve yet to see them used in basic word processing.