Exploitation and exploration: An analysis of the research pattern of Nobel laureates in Physics

As the representative of prominent scientists, understanding the Nobel laureates' research patterns throughout their careers are helpful to promote the science development. We use the BERT model to vectorize the laureates' papers and calculate the similarity matrices among them, and detect the research topics of each laureate by Affinity Propagation clustering. We further propose relevant indexes based on Kuhn's 'essential tension' hypothesis and divide the laureates' research topics into the Prize-winning topic, the topic semantically closest to the Prize-winning topic, and other non-Prize-winning topics. The empirical analysis of 117 Nobel laureates in Physics shows that they have a similar research pattern, that is, they tend to explore 2 to 3 topics alternately in different periods, but they usually identify the core research topics early and mainly focus on exploiting them throughout their careers, and other non-Prize-winning topics they explore are often related in some way to the Prize-winning ones.