Research And Application Of Achievement Patent Recommendation Method Based On Knowledge Mapping

With the development of science and technology and the rapid growth of China's scientific research strength,it is in urgent need of transforming scientific research achievements into social productivity.Faced with the increasing number of patents and patent authorization,how to make users find the right patents in the mass of patent information is a recommendation problem.Due to the existing recommendation algorithm,facing the cold start problem and the low recommendation accuracy,it is difficult to guarantee the final recommendation effect.Therefore,how to avoid the accuracy of recommendation and cold start is one of the urgent problems to be solved.Under the background of this research,this paper proposes a method of recommendation of achievement patent based on knowledge map,and puts forward a recommendation idea for the transformation of achievement patent,so as to promote the large-scale application of achievement transformation and technology transfer.In a word,this paper has the following three contributions:(1)Constructed patent knowledge atlas and user knowledge atlas for the patent field.In this paper,based on relational database,a knowledge graph construction method based on graph database Neo4 J is proposed.Firstly,by studying the knowledge extraction based on relational database,the defects of constructing the knowledge graph based on relational database were found,and then the knowledge graph was constructed based on the graph database,which represented the knowledge in the patent field accurately and flexibly.(2)Aiming at the problem of knowledge recommendation in the patent field,this paper proposes a patent recommendation method based on knowledge map.Firstly,the knowledge graph Representation Learning is used to obtain the user feature vector and the patent feature vector,and then the recommendation list is generated based on Learning to Rank in machine learning as the sorting model.Experiments show that compared with the collaborative filtering algorithm,the proposed method improves the comprehensive evaluation index of the recommendation effect by 20.5 percentage points,the average accuracy by 10 percentage points,and has higher recommendation accuracy.(3)Finally,under the guidance of the recommended methods proposed in this paper,the patent listing trading platform is designed and realized.This platform applies the research results patent recommendation method to the actual Web service,and realizes the personalized recommendation function of the results patent based on the knowledge graph,as well as the business function of listing the results patent of Hubei Technology Exchange.