On Classification and Extraction of Deep Knowledge in Patents Based on TRIZ Theory

In order to acquire and utilize patents which stored in the traditional patent database more efficiently and help designers get inspiration in the design of product innovation, a method of automatic patents classification based on SVM algorithm was presented and artificial deep knowledge template was constructed to assist designers to extract deep knowledge in this paper. The SVM-based classifier representative model was structured by artificial combined with probability statistics method to extract text features, which is based on the theory of inventive problem solving (TRIZ). The method was verified effectively by developing and testing on the CSharp.NET platform. Simultaneously, the template for innovation principles is able to inspire designers effectively to extract deep knowledge. The results show that the SVM algorithm which is applied to automatic TRIZ-based patent classification combined with artificial auxiliary extraction method is feasible for deep knowledge acquisition.