Research On Construction And Visual Analysis Of Commodity Knowledge Graph Based On Named Entity Recognition

In recent years,with the increasing development of e-commerce platforms,the amount of data generated has gradually become huge.How to extract effective information from heterogeneous data becomes an urgent problem to be solved.E-commerce platforms mostly use the retrieval mechanism based on keyword matching for commodity search and the matching range is often limited to the commodity title.In order to improve the retrieval rate of their own sales of goods,merchants often add text such as description of commodity characteristics or commodity preference information to the commodity title.This brings great difficulty to the identification of real names of goods.Therefore,this paper will research from the following aspects.First of all,aiming at the difficult problem of named entity recognition in the field of e-commerce,the Seq2 Seq model is used to replace the traditional recurrent neural network.Text features are extracted by neural network Transformer based on self-attention mechanism.Conditional Random Field is used to classify text features and determine their categories.A named entity recognition model based on Transformer+CRF is proposed to realizes the efficient and accurate identification of various categories of entities contained in the commodity title.Secondly,aiming at the complex problem of data storage and management in the field of e-commerce,taking named entity recognition algorithm as the core,the construction steps of knowledge map are designed.Through the definition of ontology structure,information extraction,knowledge fusion,knowledge processing and knowledge storage,the high-quality construction of knowledge graphs in the e-commerce domain has been completed.Thirdly,aiming at the problem that data analysis of knowledge map relies on strong professional knowledge,a visual analysis system of commodity knowledge graph is designed and implemented.The whole system adopts the layout of multiple figure combination.An interactive approach of overview and detail is used to visualize the knowledge graph at different scales,and the huge amount of hidden data is mined and explored.Besides,the practical application of this system is demonstrated by taking commodity retrieval and commodity portraits as examples.Finally,the experimental platform of named entity recognition model is designed to analyze and verify the correctness and effectiveness of the proposed method.By displaying the abnormalities found in the visual analysis,the effectiveness of the visual analysis system is verified,and it also reflects the practicality of the commodity knowledge graph constructed in this article.