Header photo


Literature Based Discovery

As research fields become more specialised researchers tend to interact more with people and literature from their chosen speciality and less with research outside of their own specific area of interest.

Ian Ruthven | Mark Baillie | Ulises Cervino Beresi

Consequently, the interaction between fields, through cross-referencing across fields and shared use of common literature, becomes reduced and related fields detach from one another. The result is relatively isolated and highly specialised bodies of literature, a phenomenon that has recently accelerated due to the increased rate of new publications available online.

This detachment of research fields means that academics who share common interests and approaches but who work in different areas can miss important connections. It is becoming increasingly challenging for most researchers, especially in established fields, to keep up-to-date with important developments in their own chosen speciality. However, keeping track of useful new developments in allied fields is even more demanding, relying far more heavily on the inefficient processes of manual literature searching and browsing, chance discoveries through personal communications or selective manual dissemination of information. More often, cross-disciplinary connections have to be engineered through dedicated, but often small-scale, initiatives.

Literature Based Discovery (LBD) is a research area which aims to help discover these connections between seemingly unrelated disciplines by mining publicly available academic literature. However, LBD systems are notoriously difficult to evaluate - how can we tell when we have found useful information from an unfamiliar field?

At Strathclyde we are working on evaluation approaches to LBD by studying how people use LBD systems to uncover new information. Working with pharmacists, information management specialists and computer scientists we are exploring how people from different disciplines assess new information from other fields. Our initial findings indicate that the quality of information itself is not the only thing people care about - the presentation of new information is also important as is the type of information presented (from example, computer scientists like hard facts and evidence of success but show less preference for conceptual work). Our findings are being developed into new approaches to evaluation for assessing the quality of LBD systems.