<p>Opinion leaders refer to users who provide information to other users in social networks and, at the same time, impact other users' ideological concepts and have an essential influence on the orientation of public social opinion. In the most research literature, opinion leader mining methods mainly include two types of methods based on text contents and network features. However, the above methods still have some problems: text contents and network features are not fully integrated, and the determination of the number of opinion leaders in social networks is subjective. This paper proposes an opinion leader mining method based on text contents and network features to solve the above problems. This method integrates text contents and network features. First, it uses the text content to construct a directed graph of social networks. Secondly, it uses the K-mean clustering algorithm to identify significant communities. Then, it uses network features to measure user leadership. Finally, it uses user coverage to select users with higher user leadership from significant communities to form a set of opinion leaders. The experiment results indicate that the proposed method improve the evaluation indicator compared with the other two opinion leader mining methods. In conclusion, the method proposed in this paper can more accurately and objectively mine the set of opinion leaders in social networks.</p>