Research And Implementation Of Patent Mining Methods With Distinguishable Validity

With the rapid development of science and technology,enterprises and research institutions also have attached great importance to patent applications and produced a large number of patent documents.These patent documents provide a potential for further analysis of technological development trends.Patent documents are effective carriers of technical information.The analysis of patent documents cain obtain important technical details,leading business trends and new industrial solutions.Excellent patent mining is beneficial for enterprises to maintain their legal rights,avoid losses and even increase commercial profits,and reduce the risk of patent infringement.Because the total number of patents is very large,the time complexity of mining is larger.Particularly,xwhen the descriptive statements of patents are longer in length and larger in quantity and involve more technical terninology,it is found that the process of patent evolution is not effective.In addition,the majority of existing comparative patent mining methods lay emphasis on the dif'ferences between patents and cannot represent the relationship between patents very well.To solve the above problems,this paper mainly studies the related issues of patent documents comparison.The resulting summary can help the patent analysts to quickly understand the relevance between the pending patent and patent database and detect the substitution of patents or the patentability of new patent applications.The work of this paper mainly includes the following aspects:(l)In order to solve the problem that a large number of patents need to be compared,which leads to a high time complexity,a patent evolution modeling method based on multi-point graph is proposed,which can generate the evolution tree of related technology field for a given set of patents,and select representative patent set.Thus,in the follow-up study,it avoids comparing with less relevant patents in the patents set,and the time cost is reduced.The experiment results show the effectiveness of this method.(2)Because the relationship between patents cannot be visually seen among the existing comparative summaries,and the results cannot be adjusted according to the different needs of users,this paper proposes documents retrieval modeling method based on comparative patents.This method first extracts the items that can be discriminated from each patent document,then connects these items in the same graph through methods based on graph theory.And by using user-defined stop words,it can quickly get the result that the user wants.The experiment results show that,compared with the existing methods,this method can get a comparative summary more accurately.(3)This paper proposes a distinguishable patent mining method with validity,which can be used to generate a comparative summary between the pending patent and the entire patent database.Based on the field of the pending patent,this method first used patent evolution modeling based on multi-point graph to construct the representative patent set.Then,this study adopted the method based on comparative patent literature retrieval modeling to construct comparative abstracts for the representative patent set.The experiment results show that this method reduces the time cost of patent analysis.