[Purpose/significance] The purpose of this paper is to improve the resolvability of sentiment analysis results of network users’ review by the construction,projection and visualization of bipartite network. [Method/process] This paper constructs the bipartite network of"user-product feature",and combines with bipartite network projection and sentiment analysis results to get the "product feature sentiment one mode network"and "user one mode network". And the centrality analysis and average clustering coefficient analysis of the above two networks are carried out respectively. [Result/conclusion] The sentiment differences between products under the bipartite network structure,comparison of sentiment tendencies of in different products,and the co-occurrence and relevance among different products feature groups are discovered. The paper proposes a method of sentiment analysis from the perspective of bipartite network,which improves the effect of network users’ comments mining. [Limitations] The selection of research objects needs to be extended to other areas.