Working Structure Knowledge Acquisition from Mechanical Product Patent Based on Natural Language Understanding

According to the increasing demand of patent knowledge in product innovative design, a working structure knowledge acquisition method, from mechanical product patent, was presented. A representing model of concept graph (CG) for patent working structure scheme, composed of technical component, technical relationship and semantic role, was established, which described the concept semantic hierarchy of patent technical component and technical relationship in mechanical field. This model divided CG into product structure composition CG, structure attribute CG and behavior CG. Furthermore, natural language understanding process of product working structure knowledge was put forward. The key technique of syntactic analysis and semantic analysis was illuminated. This method could effectively accelerate processing patent texts and establish a good foundation for sufficient utilization of patent knowledge in product innovation.