Topic segmentation is the task of dividing unstructured text into thematically connected segments (such as those dealing with the same matter). A knowledge graph is a graph structure whose vertices are various objects and the edges are the relations between them. Both the task of topic segmentation and the task of automatically constructing a knowledge graph are not new; therefore, there are many algorithms for solving them. However, methods for solving the problem of topic segmentation using knowledge graphs have so far been rarely studied. Moreover, it still cannot be said that the problem of topic segmentation has been solved in a general way, i.e., there are algorithms that, if properly configured, can solve the problem with the required quality on a specific data set. In this paper, a new method for solving the problem of topic segmentation based on knowledge graphs is proposed. The use of knowledge graphs in segmentation allows us to use more information about words in the text: in addition to being based on cooccurrence and semantic distances (like classical algorithms), knowledge graph-based methods can apply the distance between words on the graph, thereby incorporating factual information from the knowledge graph into the decision-making process of partitioning the text into segments.