A major challenge for next generation data mining systems is creative knowledge discovery from diverse and distributed data/knowledge sources. In this task, an important challenge is information fusion of diverse representations into a unique data/knowledge format. This paper focuses on the graph representation of data/knowledge generated from text documents available on the web. The problem addressed is how to efficiently and effectively create an information network, named a BisoNet, from large text corpora. Several options concerning node and arc representation are discussed, and a case study information network is created from articles concerning autism, downloaded from the PubMed repository of medical publications. Open issues and lessons learned concerning representation choices are discussed.