Research On Patent Multi-to-one Negation Relation Model

Since the 14 th Five-Year Plan,China’s intellectual property development has achieved remarkable results,and with the innovation-driven development strategy continuously implement,the awareness of intellectual property in the whole society has also been greatly improved.The patent is an important part of intellectual property,and the negative relationship between patents has also received widespread attention.At present,driven by various related strategies and policies,the number of patents is increasing yearly,and the negation relation between patents has become more intertwined,which makes it more difficult to construct the negation relation between patents in manual patent examination.Therefore,how to automatically construct the negation relation between patents has become a hot research topic of patents.Constructing the negation relation between patents includes not only the negation of one-to-one patents,but also the negation of multi-to-one patents.But most of the current research is about one-to-one negation of patents,and there is almost no relevant research on multi-to-one negation of patents.Therefore,the current research on patent multi-to-one negation relation is still in the initial stage,and there are still some challenging issues to be solved,including: how to define clearly technical features in the comparison process of negation relation,how to play the role of patent technical feature function and effect role,and how to implement the feature combination of multiple patents.To solve those challenges,from the technical perspectives of supervised learning and unsupervised learning,we use the technical methods of game theory and convolutional neural network to research the patent multi-to-one negation relation model.(1)Unsupervised learning: We propose a patent multi-to-one negation relation model based on game theory.First,we clearly define the technical features of the patent as the technical utility of the patent and the technical function-effect of the patent.To be more specific,the technical utility of the patent refers to the functions and effects that can be achieved by the technical features of the patent.The technical function-effect of the patent represents the relationship between the function and effect of the patent technical feature,so as to fully exploit the function and effect of the patent technical features.Then,in order to maximize patent technical feature function and effect in the comparison process of patent negation relation,we construct the patent multi-to-one technical utility game tree and the patent multi-toone technical function-effect game tree based on technical features of patents.Finally,we realize the judgment of patent multi-to-one negative relationship and the prediction of negation probability through the game.(2)Supervised learning: We propose a patent multi-to-one negation relation model based on convolutional neural network.First,we use the above-mentioned negative probability of unsupervised learning as the label of this model.Second,patents are represented by technical utility and technical function-effect,respectively,so as to realize patent technical feature function and effect to maximize their role in the determination of patent negation relation.Then,they are sent to the convolutional neural network,and the operation of the convolutional neural network is used to realize the combination of technical features of multiple patents.The model is trained,and the judgment of patent multi-to-one negative probability and negative relationship is finally obtained.Based on the game theory and convolutional neural network,this thesis studies the patent multi-to-one negation relation model.Using the patent technical functioneffect and technical utility as the technical features,it explores the negation relationship of multiple patents to one patent,which lays a foundation for the subsequent construction of patent negation network and the calculation of patent portfolio value and strategic value.