Semantic analysis is an important part of natural language processing, and it has a broadly application in the network information processing. This paper presents a semantic analysis framework based on the directed weighted graph. The paper uses a directed weighted graph for semantic classification. The methodology takes the information semantic analysis as the goal in network-oriented approach, particularly in E-commerce user reviews. It looks the formal semantics lexical as semantic body and denoted by nodes. It uses links to represent relationship between the nodes and calculated by weights. Directed links are used to connect to each other in nodes, which semantic vocabulary is interrelated between nodes. Then a directed weighted graph is constructed by semantic nodes and their interrelationships relations. The experimental results and analysis show that the proposed method in the paper can classify the semantics into different classification according to the path length threshold requirement.