Research On Patent Automatic Classification Method Based On TRIZ

Patents are the information resources with the highest content of new knowledge and contain the most heuristic principles.Patent information also contains many solutions to technical problems.However,with the in-depth study of patent classification methods and text mining,the traditional patent classification method can only retrieve the technical field,application field and its functions.Traditional classification methods only provide limited knowledge in a certain field,and many new technologies are usually interdisciplinary,which can't excavate potential and more valuable technical information from patents,such as creative thinking methods and creative techniques to solve technical problems,so they can't be used as an auxiliary product innovation tool.For developers who want to innovate their products,it is not advisable to classify Patents according to the traditional classification method.In recent years,many scholars have tried to classify Patent Texts Based on the theory of TRIZ(teoriya resheniya izobreatelskikh zadatch)in order to find creative solutions to solve technical problems.In order to dig deeply and make full use of the innovative ideas and key technologies in existing patent documents,this paper designs an automatic classification scheme of Chinese patent documents based on the invention principle of TRIZ theory.In this scheme,a TRIZ dictionary of invention principles is constructed based on the classical description of TRIZ invention principles and the patent documents labeled with the invention principles.The dictionary of invention principles is used to segment and select the features of patent texts.Machine learning method is used to train and predict the classification model of patent samples.In order to speed up the scheme verification,the construction of TRIZ dictionary of Invention Principles,the word segmentation of patent text based on TRIZ dictionary of invention principles,feature selection and feature vectorization are all implemented by software,and the training and prediction of classification model are realized by support vector machine.The experimental results show that the scheme can achieve high classification accuracy.Therefore,the implementation of this method can provide a reference for researchers of automatic patent classification based on TRIZ invention principle.In order to let more people understand TRIZ theory and experience the process of patent classification based on TRIZ invention principle,this paper designs and constructs a patent classification case base system based on TRIZ invention principle.With this system,users can not only upload patents,but also perform word segmentation,feature extraction,classification and other operations on patents.Finally,the user's classification results of patents are successfully stored in the case base after the administrator's authentication and assessment.