Predicting Topic Evolution in CQA Sites usingTemporal Analysis

Community Question Answering (CQA) Sites are online knowledge exchange platforms that are a warehouse of information. They cater to a variety of domains by allowing users to post questions and answers specific to the domain. CQA sites are a crucial source of information for keeping up with the rapidly changing technologies. As new technologies are launched in their respective domains, these sites experience a heavy inflow of posts related to these cutting-edge technologies. People keen on keeping up with the new technologies in their fields can heavily benefit from CQA Sites. Since these sites are loaded with posts related to trending as well as evergreen topics, it becomes difficult to isolate the trending topics from the obsolete ones. In such a situation, a temporal analysis of CQA sites could be of great help in identifying trending topics. In this work, we consider the dynamic nature of CQA sites and use temporal analysis to not only detect the changes in topics discussed in these sites but also look at the social interactions within the discussed topics at a micro level. Social interactions determine how well a social community is performing and also its future performance. CQA sites have different types of social interactions. In this thesis, we compare three time series prediction algorithms for predicting eight types of social interactions. Our analysis is done on a real-world dataset from Stack Exchange CQA site.