The ‘Open Discovery’ Challenge

Abstract

One of the most exciting goals of literature-based discovery is the inference of new, previously undocumented relationships based upon an analysis of known relationships. Human ability to read and assimilate scientific information has long lagged the rate by which new information is produced, and the rapid accumulation of published literature has exacerbated this problem further. The idea that a computer could begin to take over part of the hypothesis formation process that has long been solely within the domain of human reason has been met with both skepticism and excitement, both of which are fully merited. Conceptually, it has already been demonstrated in several studies that a computational approach to literature analysis can lead to the generation of novel and fruitful hypotheses. The biggest barriers to progress in this field are technical in nature, dealing mostly with the shortcomings that computers have relative to humans in understanding the nature, importance and implications of relationships found in the literature. This chapter will discuss where current efforts have brought us in solving the open-discovery problem, and what barriers are limiting further progress.