As we know, the wordnet(online dictionary from princeton university) db is organized in "synsets",so each word has several synonyms, and those synonyms in turn have their own synonyms, and so on. This arrangement can be used to dynamically generate a multiple choice question for each word. Here this will pick a random word from a attached file and generate a multiple choice question having five answer choices, of which one choice alone is correct (ie it is one of the valid synonyms of the given word) and the other four choices are randomly picked from entries in the db. So far, so good. But there is a problem: What if, among the four "wrong" choices, there are words close to the main word? Maybe another direct synonym, or the synonym of a synonym? Thus we should define a metric, "Minimal Path Length" or MPL that tells us "how far apart two words are, in their meanings". For Eg, the word "be" has a synonym "live", and "live" in turn has a synonym "endure", which in turn has a synonym, "suffer". So the words <"be", "suffer"> have an MPL of 3 as you have to hop 3 synonyms to get to one from the other. Of course, this is assuming that there are no common synonyms across this set (eg. "suffer" is not a synonym of "live") that's what "Minimum" means, ie it's the shortest path. So, use the above logic to ensure that your four wrong answer choices are at an MPL >= 4 from the main word. Now, for each question, the generated answer choices have 1 right choice and 4 very wrong choices. But an intelligent question will also have a "distracter" answer choice, ie one that's close to the right answer but is not the right answer. Ie 1 right choice, 1 distracter choice and 3 very wrong choices.