Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

To my astonishment, this writeup was actually interesting. I wonder what they could do to improve the quality of the comments on TechCrunch. These 11 types seem to make up the vast majority of the things said there, unfortunately.


People usually are lazy to express their agreement with someone, however when we have different views we want to let our opinion be known to the world ASAP.

That's why usually the louder sectors of users/readers/etc are the ones that are not pleased for one reason or another (or the most bored ones). It's very obvious in customer support forums (take for example a MMO forum, 90% of the posts are whines, although it only represents a 20%? of the real playerbase).

Probably the only thing to do to improve quality is write what they want to hear, but then you'd have the group that was previously happy complaining about how you betrayed your own standards, etc. And why should you submit to them and change your ways, in any case?

IMO Techcrunch is fine as it is, it's people who need to shape up and have some manners in 'teh internetz'. Remember that PennyArcade comic? :P


I'm all for an optional up/down vote with more granularity (with one click.) There are ambiguous situations where you get stuck on just voting up or down. For example, you might disagree with a post but consider it important to the discussion. Another typical problem is the +1/bump vote on forums. And "Thank you." A whole thing is the flag for relevancy. It gets messy, but IMHO something is needed.


Looks like a multiclass text classification problem to me. Adaptive Dynamic Markov Chains or Predictive Partial Matching combined with Error-Correcting Output Codes should handle it pretty well.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: