I am working a approximate string search algorithm. I am wondering how to go with validating that an algorithm is better than another.
I can not come up with validation set, since I have no example of misspelled words. Should I synthetically construct misspelled words and compute precision and recall over that?
The dataset is small enough that I can do a full search all the words against all other words. What metric should I use to tell what approach gives best results?
