Direction awareness in citation recommendation

Literature search is an important part of academic research. The increase in the number of published papers each year makes manual search inefficient, hence, automatic methods must be devised. Unfortunately, traditional search engines use keyword-based approaches to solve the search problem which are prone to ambiguity and synonymity. This paper focuses on the problem of extending a set of references using the citation relations between the documents. In particular, we introduce the class of direction-aware algorithms which weight the importance of incoming and outgoing edges of the citation graph differently based on user preferences. Using such an algorithm, the user can easily focus her search to-ward recent developments or traditional papers. We present two direction-aware algorithms and show that they are bet-ter suited at solving the problem at hand than state-of-the-art recommendation methods. One of these algorithms is currently deployed in a publicly available web-service called theadvisor. 1