Predicting Collaborative Edits of Questions and Answers in Online Q&A Sites

Abstract

Collaborative editing can play an important role in online Q&A sites, including iteratively advancing solutions and significantly improving the quality of questions and answers. However, the value of collaborative editing has not been fully utilized. Currently, there is no way for users to easily distinguish questions and answers which need be collaboratively edited from other ones in many online Q&A sites. For example, in Stack Overflow, there is no indicator to tell users whether the question/answer being seen need be edited or not. Thus, to make better use of collaborative editing, in this paper, we propose a framework to predict whether questions and answers need be collaboratively edited just after they are posted. The framework mainly extracts features from questions, answers, and posters (of questions and answers), and adopts machine learning techniques (e.g., LDA, SVM) to do prediction. To evaluate the framework, we chose Stack Overflow as our study platform and conducted experiments with millions of questions and answers. The results show that the proposed framework can achieve very high accuracy and be efficiently adopted in different online Q&A sites.