I was asking a question in serverfault.com and this message warned me (correctly!):
The question you're asking appears subjective and is likely to be closed
So, I'm curious if there any classifier or machine learning technique behind, or if it is just some heuristics on keywords.
Do you know?
Don't use the following words in a subject title - it connotes subjective question.
Jeff posted the regex in uservoice (As of May 2, 2009):
Regex.IsMatch(s, @"\bbest\b|\bworst\b|\bhardest\b|\byour?\b|\bfavou?rite\b", RegexOptions.IgnoreCase);
It is important to note that avoiding the above words does not make your question any less subjective - "Best management tool" is inherently subjective - the warning is not there to tell you to avoid certain words, but to avoid the question altogether, or to re-word it so it can be evaluated objectively and measurably so that the answer is the same for everyone in your situation.
This has also been deployed at StackOverflow.
if (post.Contains("serverfault") ||
post.PhraseLike("what is the [funniest|worst|best]*")) {
// looks like meta question
IsSubjective = true;
return SubjectiveMessageResult();
}
Possibly the best way to approach this problem would be using a Bayesian classifier; the same technology that is used for spam. As the site has a good training data set of questions that have and/or have not been closed as subjective, it should be able to determine with a high probability whether a new question is subjective, in exactly the same way as a spam filter does with spam.
Whether they actually use this technique, I don't know.
Keywords. It's matching the titles against a list.
I think there should be a special case for situations in which best is contained in "best practice"
