Towards identifying unresolved discussions in student online forums

Online discussion is a popular form of web-based computer-mediated communication and is a dominant medium for cyber communities in areas of information sharing, customer support and distributed education. Automatic tools for analyzing online discussions are highly desirable for better information management and assistance. For example, a summary of student Q&A discussions or unresolved questions can help the instructor assess student dialogue efficiently, which can lead to better instructor guidance for student learning by discussion. This paper presents an approach for classifying student discussions according to a set of discourse structures, and identifying discussions with confusion or unanswered questions. Inspired by the existing spoken dialogue analysis approaches, we first define a set of forum speech acts (F-SAs) that represent roles that individual messages play in threaded Q&A discussions, such as questions, raising issues, and answers. We then model discourse structures in discussion threads using the F-SAs, such as whether a question was replied to with an answer. Finally, we use such discourse structures in classifying and identifying discussions with unanswered questions or unresolved issues. We performed an analysis of the discussion thread classifiers and the system showed accuracies from 0.79 to 0.87 on several discussion classification problems. This analysis of human conversation via online discussions provides a basis for development of future information extraction and intelligent assistance techniques for online discussions.