Research And Implementation Of Construction Method For Internet Of Things Capability Knowledge Graph Based On Information Extraction

Internet of Things Capability Knowledge Graph(IoTCKG)integrates the characteristics of knowledge map and Internet of Things The IoTCKG is constructed by extracting the functions of Internet of Things and their relationships,so that the service in Internet of Things is structured and semantic information is added to the service.It is finer in granularity than the traditional knowledge graph of Internet of Things objects.The purpose of the IoTCKG description method is to solve the description problem of the Internet of Things,which can be used for the integration of heterogeneous Internet of Things platforms,service discovery and recommendation systems,etc.Because heterogeneous Internet of Things standards and descriptions are not conducive to integration and further innovation in the field,in recent years,the traditional Internet of Things(IoT)has transited to the emerging WoT(Web of Things)to solve some of the problems of the Internet of Things,but most of them are to build knowledge graph of Internet of Things objects,which have relatively large descriptive granularity and can not meet the dynamic changes of object functions and higher-level application requirements.Based on this background,this topic innovates the composition of knowledge graph by constructing the capability knowledge graph of the Internet of Things,the traditional granularity of "thing-relationship-attribute" is refined into "capability-relationship-capability" capability knowledge graph to describe the functions of objects and their relationships in the Internet of Things.Meanwhile,the semantics of the Internet of Things is added.To this end,this paper has done the following specific research and work:Firstly,the problems and unique characteristics of the Internet of Things are studied and described,and a scheme to solve these problems by using the capability knowledge graph of the Internet of Things is put forward.On this basis,a set of solutions to construct the capability knowledge graph of the Internet of Things are put forward.Secondly,it studies how to extract competence knowledge vocabulary from corpus.According to the characteristics of the Internet of Things,a BiLSTM model is proposed to extract target competence vocabulary from segmented corpus and provide node data for constructing knowledge graph.Thirdly,it studies how to extract relations from sentences marked with competence pairs,and studies supervised and semi-supervised information extraction.In the supervisory field,BiLSTM-CNN model is proposed,which combines the advantages of LSTM and CNN.In the semi-supervised domain,a RETAG model is proposed,which can obtain more competent vocabulary information and better relationship extraction results.