Closed Bug 660135 Opened 15 years ago Closed 13 years ago

Evaluate if a question is easy to answer

Categories

(support.mozilla.org :: Questions, task)

task
Not set
normal

Tracking

(Not tracked)

RESOLVED WONTFIX

People

(Reporter: stas, Unassigned)

Details

Using machine learning (even the most naive classifiers out there), it should be possible to look at a question and estimate if it's easy to answer or not (I'm using a binary scale here, not 'how easy it is', since it simpler to implement). This data could be used in two ways: - show a list of 'easy' questions to new contributors to get them started, - learn which keywords in the question contribute to the fact that a question is classified as 'easy'. To train the classifier, we could either: - have a button that the support contributor hits after answering the question, which says 'that was easy to answer,' or - have a similar button that says 'that was *hard* to answer,' or - just see which questions have been answered (?)
Huh, this sounds like fun. Even a naive Bayesian should be a start at this.
Looks like using a library like https://code.google.com/p/django-spambayes/ could be enough. In this case, we wouldn't be detecting "spam", but "how easy the question is." For the classifier, it's the same thing. I actually started some coding this evening, but I might need some help with setting up kitsune locally... :(
I don't think this is still an issue at present because of the resources we now have but maybe we should have a discussion on this then come back here so we can get a kick on it?
Flags: needinfo?(stas)
I think we should wontfix this bug now. I don't have much time to experiment with this anymore.
Status: NEW → RESOLVED
Closed: 13 years ago
Flags: needinfo?(stas)
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.