Learning to rank experts using combination of multiple features of expertise

In academic domain, technical conferences are conducted to share different research ideas and to propose new research methodologies. The number of conferences being conducted in different academic domains and the number of research participants in conference are increasing rapidly. The conference chairs face difficulty in assigning panel of reviewers for various research topics. A ranked list of experts in a specific topic would assist the conference chairs in finding panel of reviewers. Expert finding system provides solution to this problem. The task of expert finding system is to determine a list of people sorted by their level of expertise in a specific research topic. This paper combines multiple features of research expertise to rank the list of experts on a topic. The ranked list of experts can be used to update the topic relevance score of a researcher for a specific research area. In order to rank the experts, we have used novel time weighted Citation Graph based features, modified Latent Dirichlet Allocation based textual features and Profile based features to represent the expertise of a researcher. Rank aggregation is done based on multiple features. Lambda rank, a semi-supervised learning to rank algorithm is used to learn the ranking function. The ranked list of experts for a research topic is suggested using the learned ranking function. Experiments made over a dataset of academic publications in the area of Computer Science using combination of new features of expertise provide better ranked list of experts than using individual features.