Serving the readers of scholarly documents: A grand challenge for the introspective digital library

The scholarly literature produced by human civilization will soon be considered small data, able to be portably conveyed by the network and carried on personal machines. This semi-structured text centric knowledge base is a focus of attention for scholars, as the wealth of facts, facets and connections in scholarly documents are large. Such machine analysis can derive insights that can inform policy makers, academic and industrial management, as well as scholars as authors themselves. There is another underserved community of scholarly document users that has been overlooked: the readers themselves. I call for the community to put more efforts towards supporting our own scholars (especially beginning scholars, new to the research process) with automation from information retrieval and natural language processing. Techniques that mine information from within the full text of a document could be used to introspect a digital library's materials, inferring better search metadata, improving scholarly document recommendation, and aiding the understanding of the text, figures, presentations and citations of our scholarly literature. Such an introspective digital library will enable scholars to assemble an understanding of other scholars' work more efficiently, and provide downstream machine reading applications with input for their analytics.