ABSTRACT
In this chapter, the authors provide details about a few use cases where graph modeling of the task provides us with the flexibility to adapt existing techniques with ease on large-scale data. They introduce an interesting and very important problem of hypotheses generation in biomedical text mining and a relatively new hybrid approach that leverages and combines principles from the information retrieval (IR) field, and graph traversal to generate high quality, reliable postulates. Guilt by association is a common approach used in social network analyses, where new associations with a node in the graph are determined based on its existing associations and the associations of candidate nodes. This principle has been widely adopted in the biological domain. Initially formulated as an IR problem, literature-based discovery has grown to be one of the foremost tasks in text mining in the biological domain.
