Serendipity in context: Prioritised contextual browsing in large‐scale digital libraries

ABSTRACT In previous work we proposed a model of information‐seeking behaviour in scholarly workset creation, combining and extending established models by Bates, Ellis, and Wilson to encompass strategies for scholarly research in large‐scale information systems. However, this model simplifies contextual browsing, a key aspect of information seeking in large‐scale information systems, as a single, holistic, strategy. Here, we extend this model with granular strategies for contextual browsing, defining new modes to characterise contextual browsing as combinations of these strategies, which we show to be consistent with serendipitous discovery as described by Makri et al. We study the properties of prioritised contextual browsing as a mechanism for implementing these strategies. We describe the Compage framework, a proof‐of‐concept implementation for prioritised contextual browsing of Linked Data resources, using Jaccard similarity for prioritisation. Extending Compage, we develop a simulation environment in which we investigate the utility of prioritised contextual browsing over a large‐scale digital library dataset. Our simulation applies three strategies for the traversal of contextual metadata: reset, unprioritised, and prioritised. Results empirically demonstrate the advantages of prioritised contextual browsing, and that elements of serendipity can be identified and incorporated within our information‐seeking model. In doing so, we evaluate our model's suitability for this scenario, yielding a more detailed understanding of the strategies and modes of behaviour underlying contextual browsing.