Research On Method Of Patent Knowledge Acquisition And Its Pushing

After comprehensive analysis of topical web crawler,patent text classification,potential design requirements and product innovation design method based on patent knowledge,this paper proposes the method of design and implementation of the patent topical web crawler system,semi-supervised patent text classification method based on improved Tri-training algorithm and customized patent knowledge push method,and developed the Computer Aided Product Innovation Design Platform Based on Patent knowledge(CAPIDPPK),in order to solve the insufficiency both existing in current patent data acquisition and patent text classification,and a series of problems existing in the process of patent knowledge push.There are five chapters in this thesis.The chapter 1 expounds the research background and meaning of the subject,mainly introducing the domestic and foreign relative research status about patent,topical web crawler,patent text classification,potential design requirements and patent knowledge aided product innovation design,so the research approach and the main research content of this thesis are determined.In the second chapter,this thesis designed and developed a patent topical web crawler in order to solve the problems of low acquisition efficiency of patent data and high proportion of manual labor.A similarity calculation method based on sentence vector(Doc2Vec)was proposed on the discrimination of patent topical relevance to conduct effective preliminary screening of patents.The results of the experiments indicate that the topical web crawler has high acquisition efficiency and applicability.In the third chapter,in order to solve the shortcomings of Information Gain(IG)' using in patent text classification and the bottleneck problem of training set labeling in traditional patent automatic classification,an improved IG algorithm and a semi-supervised classification method were proposed respectively.Finally,the thesis takes the classification experiment of American patents as an example.It indicates that the proposed classification method has positive automatic classification effect in the case of fewer labeled training samples,and the improving of the classifier' generalization ability through appropriate increasing of unlabeled sample data.The fourth chapter analyzes the shortcomings of existing patent knowledge push research,and proposes a customized patent knowledge push method.Firstly,based on the designer model and the design requirement model,the matching patent knowledge is pushed to assist the designer to carry out the first innovative design.After designer' first innovation design completed,the potential design requirement is explored to establish the potential design requirement model.So the patent knowledge meets the new reqirements is pushed once again to assist the designer to carry out the second innovative design,realizing the purpose of actively providing customized patent knowledge for the designer in the process of product innovation design.Finally,taking the innovative design of digital measuring element control box with high reliable waterproof and dustproof performance as an example,the validity and feasibility of the knowledge push method are preliminarily verified.The fifth chapter develops and implements CAPIDPPK based on the theory of the second,third and fourth chapters.It explains the implementation and using method of each functional module of the platform,so the effectiveness of the platform is preliminary.verification.The last chapter summarizes the work of the full paper and looks forward to the follow-up work.