Social question and answer system for large-scale web data

To study the social question and answer system (Social QA) for large-scale web data, we discuss some component techniques for the social QA system, such as question pre-processing (question analysis and question semantic expansion) and question matching. Experiments on a set of 15 million question and answer pairs from the web were conducted. Results show that our system achieves the question matching accuracy over 90% in the closed test and that over 70% in the open test. It can be concluded that this system is efficient and can basically meet the real-time requirements.