Fecha de publicación: 18/07/2023 Fuente: Wipo "BigData" According to the intelligent interaction session big data analysis method and the big data server provided by the invention, the structured semantic vector of the to-be-analyzed request response interaction session is obtained through the deep structured semantic model, and the risk thermodynamic description knowledge of the to-be-analyzed request response interaction session is obtained through the deep thermodynamic description knowledge model; according to the method, different session semantic details of the to-be-analyzed request-response interaction session can be extracted as much as possible, then high-quality aggregation can be performed according to the structured semantic vector and risk thermodynamic description knowledge of the to-be-analyzed request-response interaction session, and a session risk joint description item of the to-be-analyzed request-response interaction session is obtained. According to the invention, the method achieves the enrichment of the structured semantic vector, improves the precision of the risk description of the mined session text as much as possible, can carry out the precise and credible risk prediction analysis through the accurate and abundant session risk joint description items, and reduces the risk prediction analysis deviation caused by a feature mining error.