|Think about Loose Coupling|
First of all you have to build the set of suggetions.
For this purpose, you have to avoid string distance as Hamming, Text::Levenshtein or Text::WagnerFischer (but you can use them later)
Instead you can use a phonetic algorithm like Text::Soundex (that you should find along with Perl), Text::Metaphone and Text::DoubleMetaphone.
They transform a word in a phonetic code that you can use to retrive suggested words:
hullo->(phonetic alg.)->h2l3 (note: same code for phonetic similarity)
So you can made an hash on disk (e.g. with DB_File) with phonetic codes as keys and as values the words that have the same phonetic code:
and so on.
After the hash (that you have built before running the spell checker), you parse the text; you have to:
1) isolate the word
2) calculate the phonetic code of the word.
3) retrive the suggestions for this word (i.e. words that have the same phonetic code)
4) see if the word is in the set of suggestions
4a) if yes, parse another word (i.e. the word is in my corpus)
4b) if not, prompt the user THIS set of suggetions
4ba) now you can use the distance string algorithms to sort the suggested words (e.g. with Levensthein).
To retrive the words you can do for example: