Question Answerer Recommending In Question Answering Community

With the fast development of question answering communities such as Baidu Zhidao and yahoo! answer, more and more users are attracted and use them to share information and ask people to help for some problems. Due to the large amount of new generated questions, askers have to wait more and more time for the answers. Meanwhile, the answerers have to browse the question category hierarchy to find interesting questions among millions of open questions to answer. It is really time-consuming and enthusiasm-dispelling even with the help of organized question categories.In this paper, we focus on addressing this problem by recommending question answerers, in which a question is given as a query and a ranked list of users is returned according to the likelihood of answering the question. We proposed a two-steps framework to resolve the problem. In order to recommend question answerers, first, we need an algorithm to represent users' topic interest and expertise. In this paper, we use language model to learn users' topic interest and we propose an answer quality based framework to model user expertise. Since user model is based on history information, users may become inactively after some time, so we need to consider user availability. In this paper, we use exponential distribution to model user availability and estimate the probability of users to answer the recommend question in a given time.Experiments are carried out on a real-world data crawled from Baidu Zhidao during Jan 1 2010 to Jan 15 2010, which consists of 1017461 questions,2992870 answers and 1160723 users. The experimental results reveal significant improvements over the baseline methods and validate the positive influence of user expertise information.