25 February 2010

The ultimate in contextual spell checking

Have you noticed, how advanced (some) spell checkers have become ?

Google Wave has a contextual spelling checker, i.e. it uses the context to determine if a word is correct. I don't know the ins and outs (yet), but I imagine that Google is building up knowledge based on statistical analysis of web input (the same way it's translation tool uses statistical learning techniques).

Here are some tests of the Wave contextual spell checker ..

Their here.

Hmm, no luck, maybe too short, maybe not enough context ? It's being accepted. How about:

I am over hear.

Nope. Maybe I'm still expecting too much .. but the system is supposed to learn from input all over the web .. so what about:

Can I have too bottles of beer ?

Success ! "Too" is underlined and the option of "two" is shown. Clever. This begs the question, how this will develop with time. Google translate is already becoming very able (very quickly), so what does the future hold in store for contextual spell checking ?

No more documents, containing flawed language .. or will this cause us (humans) to become lazy when it comes to language ? The laugh is, if we do become "language lazy", the statistical methods used by Google and others will (with mass laziness) start learning this new language. So a tool to improve our language could, over time, very well change our written language ..

Well, I've devised a test, or rather a milestone, to see how well contextual spell checkers are maturing:

Sir Lancelot is a night.

(That will be the dawn of artificial intelligence, statistically learnt from webpages the world over).

No comments: