The search mechanism is obviously simple character-string match and the result includes quite a few false hits. This can be avoided by treating the search input as token rather than substring.