<p>With the rapid development of science and technology, analyzing academic trends is essential for understanding research fields and stay abreast of new developments. In the related work of academic topic analysis, little is known about how academic hot topics emerge and how the interaction between authors and topics contributes to the evolution of topics. This paper reveals the dynamic characteristics of the emergence of academic hot topics by viewing the scientific communities as mutualistic systems, where the interaction between authors and topics resembles the mutualism structure in the ecosystems. Based on data from the Artificial Intelligence (AI) domain in the Scopus database, this paper constructs author-keyword bipartite networks to explain the changes in hot topics and the interaction patterns between authors and topics over time, using the network structural metrics of nestedness and modularity. Results show that the emergence of hot topics can be detected by the coupling trends between nestedness and modularity. The observed rise of collective attention in the AI community can be characterized as a network structure transition from modular to nested.</p>