amora | semantic search engine where anyone can define anything

Recall that Levenshtein distances count the number of substitutions, additions and deletions required to transform one string into another. Because of this, they often are minimized when comparing strings of similar length (because even if a lot of substitutions are required, you don't have to add or remove a bunch of characters). You can see this playing out in your second example where your best outputs all are the same length as your search string (len("bna") == len("A&W")).

If your search string is always going to be a single word, then your idea to calculate the distance for each word in the string is a good one since each word is more likely to be a similar length to your search string. However, currently you are doing a case sensitive comparison which means that editdistance.eval('bna', 'BNA') == 3 which I'm guessing you don't want.

try: