gcopeland (7/21/2008)
I think a hash table would be a much better implementation. It would be hugely faster, and if the word doesn't hash, it would be easy to grab all the words near the hash as suggestions.
A ternary search tree might work even better (although it would be hard to develop in SQL without using CLR). In the TST words that are similar tend to "bunch" together as well. BTW, just wondering but what type of hash function would you use to get similar words to group together? If the function is too simple you're going to pay a penalty in access costs.