Abstract

The World Wide Web contains a vast corpus of information describing a variety of events, both current and historical. However, this information is not as well connected as it could be, with many stories standing on their own without links to related events. Previous research suggests that generating these links could increase the scope for users to serendipitously discover events related to the story which they are currently reading. Existing approaches to this problem have focused on connecting events within small corpora containing only a few hundred documents, and have been more concerned with treating the problem as a challenge in information retrieval, with precision and recall being the measures of success. Our approach differs by tackling this problem in the human factors domain, with user perception as an indicator of success, as opposed to strictly technical measures. The Human Centred Event Linking (HuCEL) project provides a solution to this problem which is lightweight, extendable and builds upon existing work. By extracting keywords from the main content of a news Web page, we are able to automatically generate search queries to scan the Web for related events, and display this additional information to users next to the original story. A technical evaluation indicates that users find our queries to be related to the story, demonstrating that our algorithm is producing quality keywords. A qualitative and quantitative user study of the links generated by the HuCEL platform also demonstrates that users find these associations to be related to the story under discussion.