Predicting the Usefulness of Questions in Q&A Communities: A Comparison of Classical Machine Learning and Deep Learning Approaches

Questioning and answering (Q&A) communities have become an important platform for online knowledge exchange. With a vast number of questions posted to elicit high-quality solutions as well as a large number of participants engaged in online knowledge sharing, a grand challenge for Q&A communities is thus to effectively and efficiently identify and rank useful questions. The current approach to solving this problem is either through user voting or by community moderators. However, such manual processes are limited in terms of efficiency and scalability, especially for large Q&A communities. Thus, automatically predicting the usefulness of questions has significant implications for the management of online Q&A communities. To provide guidelines for assessing the quality of online questions, this research investigates and compares various classical machine learning and deep learning methods for predicting question usefulness. A dataset collected from a large Q&A community was used to train and test those machine learning methods. The findings of this research provide important implications for both the research and practice of online Q&A communities.