

The collected data is too big to integrate them on a client computer, so this approach is normally not offered by most spell checkers. This approach needs a lot of precomputed data from a large text corpus (yes, Wikipedia is too small for that). The third approach is to create rules based on statistical information.For this to work, the spellchecker needs to look at several words at the same time to get the context instead of just looking for one word after another.

Especially people with dyslexia know about this problem. Another example is to find homophones, words that sound more or less the same, but have been used in the wrong manner. In this kind of approach, a spell checker service will detect the wrong use of the indefinite article “”, like using an” instead of “a” before a word beginning with a vowel sound.
