Scenario-based Query Of Power Marketing Based On Knowledge Graph

Power marketing refers to the behavior of power supply enterprises providing electricity products and corresponding services that satisfy consumer needs through a series of marketrelated business activities in order to achieve enterprise goals.The formulation of power marketing strategies requires timely understanding of customer,power system,and electricity sales information,which demands flexible and scenario-based queries.Based on knowledge graphs,intelligent question-answering(QA)using natural language processing is a new type of technology aimed at obtaining information from massive amounts of knowledge graph data and answering natural language questions posed by users.To implement a QA system for power marketing based on a knowledge graph,it is necessary to address how to accurately convert natural language queries into Cypher statements,recommend relevant information related to the field that users are concerned with in practical applications,and finally present query results in a natural language and visual format.This article’s main research work includes the following aspects:(1)proposing an NL2 Cypher implementation process that integrates concept graphs.Combining business needs in the field of power marketing,a universal query intent scenario is constructed.By performing named entity recognition,entity normalization,and query intent scenario classification processing on natural language queries,structured semantic information of the query is obtained,and the semantic information is input into an adapted intent query template to generate Cypher statements.This solution ensures accuracy while increasing the generality of the implementation process.(2)Based on the need to recommend scenario-based associated information for queries,a method for recommending associated information that adapts to extended scenarios is proposed.The SOM neural network algorithm is used to cluster domain concepts and attribute names in historical queries to generate extended scenarios.The domain concept of the input query is matched with the extended scenario by similarity,and multiple query statements are generated with certain rule policies to obtain associated information for the input query.(3)Based on the structural and content features of query results,a decision tree classification algorithm is established to realize data-driven view selection,and query results are presented to users in a multi-view interactive manner.(4)A power marketing QA system is constructed using a front-end and back-end separation architecture.NL2 Cypher implementation processes,associated information recommendation,and multi-view visualization technology are combined to implement intelligent QA and query result visualization functions.The practical application of the system shows that it is easy to use,rich in feedback information,and user-friendly,and it improves the support capability for power marketing decisions.