Research On Quality Assessment Of University Invention Patents Based On GRU-attention Mechanism

At present,thanks to the implementation of the intellectual property strategy,the number of patents in China is developing rapidly.Invention patents have received great attention from all walks of life because of their outstanding characteristics and remarkable technological progress.As one of the important production areas of patents,colleges and universities have also achieved remarkable results in the number of applications for invention patents.In sharp contrast to the increasing number,the quality of invention patents in colleges and universities is not satisfactory.Therefore,how to quickly and accurately assess the quality of a large number of university invention patents applied for,reduce the probability of obtaining low-quality patents from the source of patent applications,and improve the overall quality of university invention patents,is of great significance to the layout,development and even social progress of high-quality invention patents in various universities and the entire country.The patent text data structure is well-formed,the documents are long,and there are a lot of obscure professional terms.Therefore,it takes a lot of human resources to analyze the quality of the patent text manually.In order to effectively solve this problem,the traditional methods are either simple statistical analysis based on patent citation networks,or subjective evaluation of quality based on evaluation indicators.However,the citation analysis of patents has the problem of time lag,and the patents at the initial application stage cannot be evaluated,and the index evaluation method relies too much on expert knowledge to exclude subjective effects.Both methods ignore the patent documents important data information.The rapid development of deep learning technology provides a new idea for high-quality patent evaluation.In view of the characteristics of the patent text structure,this study fully considers the patent text attributes and structured data information,adopts a qualitative and quantitative method to analyze the patent data,and builds a GRU-Attention mechanism neural network model based on deep learning technology.At the same time,this study combined with expert knowledge to construct key data indicators and weights for evaluating the quality of invention patents to improve the model.The results show that the improved model has better stability and accuracy on the training set.The final part of the article selects the invention patent data in the C9 college review as a case,and uses the constructed evaluation model to predict the quality of the invention patent in the C9 college review.The prediction result is that the proportion of high-quality patents is 16.38%,the proportion of ordinary quality patents is 54.95%,and the proportion of lowquality patents is 28.67%.Through the analysis of the status of invention patents and patent prediction results of C9 colleges and universities,it is found that the current patent development of C9 colleges has the problems of focusing on quantity and light quality,low conversion rate of achievements,and obvious lack of scientific and technological innovations for D class patents.Finally,in order to promote the healthy development of invention patents in C9 universities,based on the research results and the problems found,this paper puts forward some reasonable suggestions,such as change the focus of patent and strengthen the awareness of patent quality,rational patent layout,simultaneous development of basic,and focus on key patents and manage patents scientifically based on prediction results...