Knowledge Co-Creation Value Identification Based on Double Granularity Semantic Features and Heterogeneous Networks
[Research purpose]The purpose of this paper is to optimize the identification method of knowledge co-creation value in virtu-al communities,and alleviate the problem of poor identification effect of high-value knowledge resources caused by information overload and association complexity.[Research method]Starting from the dynamic collaborative process of knowledge co-creation,a knowledge co-creation value recognition model(DGSHAN)integrating double-granularity semantics and heterogeneous networks is constructed.Firstly,BERT and Sentence-BERT are used to obtain the semantic information of double-granularity knowledge units of words and sen-tences in parallel,and then CNN and BiLSTM are introduced to differentially refine the local kernel features and dynamic time series fea-tures of collaborative knowledge.At the same time,HAN is used to deal with the heterogeneous association network,to explore the asso-ciation rules in the multi-type entities and topological structures under user interaction,and finally to integrate the dual-link characteristics of knowledge resource combination and user behavior interaction to realize the effective identification of knowledge co-creation value.[Research conclusion]Verified by Meizu Community Flyme data,the recognition accuracy,macro F1 and weighted F1 of this model are 82.61％,73.56％and 81.39％respectively,which are significantly improved compared with other baseline models.The model can ef-fectively improve the recognition effect of knowledge co-creation value.
万方数据
维普
