Research On Method Of Intelligent Mining And Applicating Mechanism Topological Knowledge From Mechanical Patent Text

Intelligent design aims to solve a large number of unstructured decision problems in the process of mechanical equipment design.A large number of data,algorithms and knowledge are needed to support the process of knowledge acquisition,expression,organization and utilization.Aiming at the problem of how to intelligently mine and represent the topological knowledge of institutions from patent texts,this paper comprehensively analyzes the research status of patent literature technical information mining,named entity recognition,entity relation extraction and innovation design by patents’knowledge.A component name extraction method based on word frequency difference algorithm and left segmentation lexicon,a component connection relation extraction method based on semantic classification prediction model are provided,and the mapping relationship between component connection relation information and mechanism topology knowledge is established.The main research progress is reflected in four aspects.(1)Analysis of difference between information expression and professional knowledge representation of institutional topology in patent texts.In order to clarify the implicit mapping between mechanical movement scheme information described in patents and professional expression of mechanism topology knowledge,patents in the field of machinery are extensively investigated.By dissecting patent claim,mode of carrying out the invention,specifications and drawings,technology characteristics of mechanical structure are deconstructed.Based on the characteristics of components’name and their connection relationships,the target text of information extraction is determined,and a knowledge mining strategy is proposed to identify moving link and kinematic pair from components’connection relationship.(2)Automatic extraction of system component name information.In order to accurately extract the components’name used to describe the technical scheme from patent text,candidate set of component name and word calculation of frequency offset(CF_m)are introduced.What’s more,a dynamically updated thesaurus for left segmentation words is constructed.Finally,method of extracting component name is proposed based on word frequency offset and thesaurus construction.The validity and accuracy of this method are verified by comparative experiments,which lays a foundation for component connection relation extraction.(3)Intelligent extraction of component connection relation information.In order to extract the component connection relation from the complex patent text semantics,a corpus of connective relation verbs and component connection relation sample data are created.An intelligent semantic classification prediction model of patent sentences is constructed based on multi-layer perceptron(MLP),and the method based on semantic classification prediction model is proposed to extract component connection relation.The effectiveness of proposed method is verified by experiments,which provides basic information sources for further acquiring mechanism topology knowledge.(4)Precipitation,expression and application of mechanism topology knowledge.In order to separate out professional mechanism topology knowledge from component connection information,the mapping relationship between them is constructed.By the use of semantic network,topology matrix and other graphical means,mechanism topology knowledge is professionally expressed.Based on the above research,prototype system mechanism topology knowledge acquisition and representation from mechanical patent(Me CPA-System)is developed.Taking the innovative design of an upright walking mechanism as an example,specific process of using this prototype system to complete mechanism topology knowledge mining and assist innovative design is described.In the end,validity and practicability of methods proposed by this paper is verified.