Levenshtein and Similar_Text Comparison

Hi everyone, I have been using the levenshtein function and similar_text function to evaluate which is the better function to use. I have read that the levenshtein function runs faster than similar_text. I haven't noticed any runtime differences but I have noticed that levenshtein seems to be more accurate. Since the levenshtein function is based on Edit Distances and similar_text is based on Decision Trees, I'm wondering when and why similar_text would be used instead of levenshtein. Thanks. -gohaku