Utilizing context in generative bayesian models for linked corpus

In an interlinked corpus of documents, the context in which a citation appears provides extra information about the cited document. However, associating terms in the context to the cited document remains an open problem. We propose a novel document generation approach that statistically incor porates the context in which a document links to another doc ument. We quantitatively show that the proposed generation scheme explains the linking phenomenon better than previous approaches. The context information along with the actual content of the document provides signicant improvements over the previous approaches for various real world evalua tion tasks such as link prediction and log-likelihood estima tion on unseen content. The proposed method is more scal able to large collection of documents compared to the previ ous approaches.