[Purpose/significance] The development of science and technology has driven the continuous growth of the number of scientific documents. Exploring the evolution characteristics of domain knowledge from different data source dimensions will help to establish a panoramic perspective on the development of domain knowledge. [Method/process] Using journal papers, conference papers, fund projects, and patents in the graphene field in the US as data sources, the BERTopic model is used to extract domain topics, track the evolution of domain knowledge from different data source dimensions, and compare the effects of knowledge evolution analysis from different dimensions. [Result/conclusion] The results show that the topic extraction results and the context of knowledge evolution based on a single data source have different emphases; domain topics based on fused data sources present a domain panorama and help to discover potential knowledge evolution patterns.