This blog is the first one in the two blogs series: the first explaining the difficulties facing a spell checker designed for the people with dyslexia and the second blog discussing developing of a word prediction writing system for those people.
Generally speaking, developing a spell checker is a relatively simple task. All you need to do is to figure out whether a word is in dictionary and if not to suggest valid words with similar spelling. This problem sounds simple, and it is indeed simple if the following conditions are met:
The correct word you are looking for has the same number of letters as the misspelled word, or the user made only one spelling mistake.
Thus, all you need to do is to find words that have one letter difference from the word the user misspelled (edit distance is equal to 1) and present them in a way easy for the user to choose from. That could be accomplished, for example, by sorting the suggested words by their frequencies of language usage.
Unfortunately, the above pattern, equal size words and a single error, does not fit the needs of dyslexics (samples of “dyslexic” writing) since in many cases people with dyslexia misspell a word with non-equal length word, and in most cases their spelling is a phonetic spelling with multiple errors.
The challenging task is to correct a misspelled word when you don’t know neither the number of letters in the real word nor the number of errors. It is clear, however, that in order to cope with this more complex task, dyslexia spelling software shall simulate human way of thinking.
To correct a badly spelled text, we read the entire text, comprehend it, mark the words that are spelled incorrectly, and suggest corrections based on the entire text context and the grammar rules. This is a very complicated path to follow for just a piece of software!
Try Ghotit and see how it works!