[Objective] Real-time and accurate understanding of the research hotspots and evolution trends in the field of information retrieval is helpful to provide reference and assistance to researchers, which is crucial for accelerating the integration of interdisciplinary and promoting the rapid application of information retrieval technology. [Method] This article uses the accepted papers of the SIGIR Annual Conference from 2008 to 2019 as the data source. First, the LDA model is used to identify and generate topics. Second, filtering irrelevant documents according to the similarity between documents and topics, and dividing documents into multiple topics by calculating document topic discrimination; Third, by constructing the evolution path of the domain topics in the time series to show the three evolution methods, including persistence, weakening and stability; finally, through the modular community, the fine-grained evolution path of a single topic is constructed to fully demonstrate the dynamic evolution process between knowledge units within the topic. [Results] The method proposed in this paper avoids the interference caused by irrelevant documents on the topics identification and evolution path, and the multi-topic division of documents helps reveal the cross-fusion between topics. The current information retrieval field is mainly user-centric; retrieval models are continuously optimized, focusing on filtering and recommendation, focusing on semantic web technology, deep learning methods are widely used, application fields such as medical and health have gradually become the focus of the information retrieval field. [Limitations] It has a certain degree of subjectivity when filtering irrelevant documents and document multi-topic division by setting a threshold.