Design And Implementation Of Patent Creative Retrieval System Based On SAO Structure

With the rapid development of technology,the competition between enterprises has gradually shifted from the market to technology,and this change has put forward higher requirements for enterprises' control over technology.Patents are the carrier of technology under the legal framework,and the protection of enterprises' patents must also be achieved through patents.The mining and research of patent texts is an effective means of technical protection.Through research patents,functions such as technical route drawing and technology trend prediction can be realized.This technical foresight has important guiding significance for the development of enterprises and countries.Therefore,how to mine effective information from patent texts is a serious challenge at present.Most traditional solutions are based on specific rules,such as patent classification based on IPC classification numbers,and such rules are often not comprehensive and accurate.Studies in recent years have found that using deep learning-based text mining techniques can often achieve better results on patent texts.This article designs and implements a patent creativity judgment system,which is applicable not only to patent examiners but also to general users.When it is necessary to judge and analyze the creativity of a patent,it is necessary to input the claims of the patent into the system,and use the search module provided by the system to retrieve related patents.Further,this paper proposes a patent similarity judgment method based on the SAO structure of the patent claims.Based on this method,the similarity of the patent to be compared and the comparative patent is calculated,and the similarity of the patent to be compared is finally calculated through an empirical formula.When calculating the structural similarity of patent SAO,the similarity of words needs to be calculated by means of dependency syntax analysis and word2 vec.The former uses some mature and excellent open source frameworks.For the acquisition of word2 vec word vectors,the system uses patent text for training.Therefore,both patent retrieval and word vector training in this system require a large number of patent texts.To this end,this system has developed a patent text data acquisition module that captures patent text information and patent PDF information.And considering that the patent data will be updated periodically,the acquisition module will also take corresponding measures to ensure the timeliness of the system data.