Automatic Classification and Analysis of Interdisciplinary Fields in Computer Sciences

In the last two decades, there have been studies claiming that science is becoming ever more interdisciplinary. However, the evidence has been anecdotal or partial. Here for the first time, we investigate a large size citation network of computer science domain with the intention to develop an automated unsupervised classification model that can efficiently distinguish the core and the interdisciplinary research fields. For this purpose, we propose four indicative features, three of these are directly related to the topological structure of the citation network, while the fourth is an external indicator based on the attractiveness of a field for the in-coming researchers. The significance of each of these features in characterizing interdisciplinary is measured independently and then systematically accumulated to build an unsupervised classification model. The result of the classification model shows two distinctive clusters that clearly distinguish core and interdisciplinary fields of computer science domain. Based on this classification, we further study the evolution dynamics at a microscopic level to show how interdisciplinarity emerges through cross-fertilization of ideas between the fields that otherwise have little overlap as they are mostly studied independently. Finally, to understand the overall impact of interdisciplinary research on the entire domain, we analyze selective citation based measurements of core and interdisciplinary fields, paper submission and acceptance statistics at top-tier conferences and the core-periphery structure of citation network, and observe an increasing impact of the interdisciplinary fields along with their steady integration with the computer science core in recent times.