Investigations into the goodness of posts in Q&A Forums—Popularity versus quality

Abstract

Q&A forums offer users a platform to gain and showcase their knowledge and facilitate intellectual sharing among community members. An incentive based mechanism is devised as a means to encourage users to contribute to questions posed in the forum by allowing various users to Upvote and Downvote posts and answers. Though this provides the needed impetus to users to contribute quality content, oftentimes the reward mechanism is more biased towards questions on popular topics such as Java, html etc. Moreover simple questions—questions whose answer can be found in most textbooks or on the web—sometimes get a large number of votes due to a number of people finding the answer with the least amount of effort from their side. The downside of this is that quality questions—questions where the user has given thought to the topic, put in effort and which would require an effort from the answerer—is seldom answered or voted up. The investigations in this paper are oriented towards separating popularity from quality. We propose an iterative technique based on link analysis algorithms to separate the quality posts from the low quality ones even if the latter have a high number of upvotes. Such a scoring can be utilized for improving search and for highlighting good posts finding answers to which would enhance the knowledge level of the community as a whole. Experimental evaluations aligning the ranks produced with expert given ranks are promising. In two out of three cases the proposed approach had higher rank correlation with those of the experts as compared to ranks inferred through Stack Overflow assigned scores. Moreover examination of individual posts which refer to common questions illustrate that the proposed techniques have the ability to push down their scores even if Stack Overflow has assigned high scores.