Enhancing the robustness of the disruption metric against noise

Measuring the novelty of scientific papers is an important research topic. If most subsequent researches of a focal paper only cite itself instead of citing its references as well, this paper could be highly disruptive as it may start a new stream of research. However, due to preferential attachment, even if a focal paper is very disruptive, the subsequent works may still cite both the focal paper and some of its highly cited references. To eliminate the noise caused by these highly cited references, we modify the disruption metric and analyze its performance and robustness. The results show that the improved method could better distinguish Nobel prize winning papers from the others. In addition, the resultant ranking is more stable against the highly cited references and random link removal on citation network.