Research on the Internal Relationship Characteristics and Their Influences of Knowledge Sharing Multilevel Network in Q&A Community

The virtual communities have become the main position for people to create and share content in today's society. It not only realizes the dissemination of knowledge and information, but also promotes the formation of the relationship between users. The traditional related studies treat all information in Internet as knowledge, which deviate from the real situation. Therefore, this paper uses text classification technology to classify the answer texts under the topic of "English learning" in the "Zhihu" Q&A community, and extract the real knowledge under the topic. On this basis, a multilevel network about answer-users' knowledge sharing is constructed, and three subgroups with different users' node degree are divided. The multilevel network exponential random graph models are used to explore the influence of local structural characteristics formed by the relationship between users on the whole multilevel network. The results show that: When the node degrees of answer-users are small and the network structure is stable, the initiative of sharing knowledge is small and the homogeneity of knowledge content is high; if there are structural holes in the network, answer-users will create an obvious clustering effect, and the heterogeneity of shared knowledge is high; for the subgroup with the largest answer-users' node degree, the relationship between users is tight and the network structure is stable, then the shared knowledge is more heterogeneous.