A patent retrieval method and system based on double classification

Abstract

As one of the most prominent carriers of knowledge, patents can provide masses of cross-domain knowledge to support the innovative design process of products. However, there is a major problem faced by most patent-aided innovation design systems, that is, identifying and providing patents related to design requirements from a multitude of cross-domain patents. To address this problem, a patent retrieval method based on double classification is proposed in this study, and a corresponding patent-aided innovative design prototype system is also developed. Firstly, the functional attributes are extracted from design requirements and expanded to different fields based on functional bases, by which the primary patent sets are constructed via classifying cross-domain patents. Secondly, technical features are mined from the primary patent sets with the topic model, which is used to classify the primary patent sets in turn and construct the secondary patent sets, hence, corresponding patents can be obtained to solve the design problems. Finally, a patent-aided prototype system is developed based on the proposed method. On the basis of the system, the hole plug of irradiated samples of nuclear reactor pressure vessel has undergone innovative design, which verifies the feasibility and practicability of both the method and system.