Research On Key Technology Of Enterprise Overseas Patent Warning Platform

In a globalized world,enterprises often encounter obstacles of patent infringement when they export their products overseas,and patent warning work is so complex and time-consuming.And there are 2 of the most important steps in the patent warning work,which is retrieval optimization and identification of suspected infringement patent,and we need to improve efficiency of these steps.In this paper,we propose an optimization algorithm of patent retrieval based on genetic programming,for constructing the retrieval effect automatically.There are 2 aspects of the retrieval construction,that key words extraction and determination of logical relations.On the one hand,segment patent texts into words and calculate information entropy of the word,get the word which have low information entropy.And extend key words using word2 vec tool kit.On the other hand,we transfer query reformulation to query hierarchy tree by using binary tree,for facilitating the subsequent crossover,mutation and other genetic operations.And find the logical relationship between the keywords,and ultimately get the retrieval whose recall and precision comply with the requirements.In addition,we construct a patent infringement detection algorithm based on graph similarity.We extract SAO structure in the patent,and based on the principle of infringement judgment,determine whether there is a subset relationship between the graph,achieve automatic identification of patent infringement.In order to verify the effectiveness of the algorithm,we have built the prototype system of patent warning platform.And we conduct a patent analysis using the relevant patent data in the field of smart home.The result shows the method can construct a good patent retrieval and find the suspected infringement patent.And through the patent infringement detection algorithm,we identify 2 suspected infringement patents,which is indeed an infringement patent.