Research On Semantic Driven Data Query And Intelligent Visualization

With the coming of the big data era,how to excavate and utilize valuable information from mass data has become an important problem concerned by people.However,the information itself has natural concealment.From the query of the data to the result,the final observation of the data is usually unable to obtain the most valuable information.Especially for relational databases,it stores a large number of domain data.At present,the access needs the user to master the query language like SQL,although it can meet the needs of the user to query the data,but it is difficult to master.In addition,as human beings have a strong visual identification ability,the topic of data visualization has attracted more and more attention from researchers,it can help people make better decisions.Although many excellent visualization systems have emerged in the field of visualization,many challenges still remain.In particular,for the part of the user's intent analysis,most of the solutions are to allow users to use established rules to describe the query language,which undoubtedly reduces the user's experience.Or the connection between the user description language and the database query service is established by service matching,but the method needs to be described separately by the service description language,which makes the final result deviate.Based on the above reasons,this paper designs and implements a semantic driven data query and intelligent visualization technology model.In this model,users can query database data by natural language and automatically recommend charts by identifying user intentions,so that users do not need to care about the underlying implementation.The main contributions of this article are as follows:(1)learning word embedding: This paper uses the LSTM neural network as the basis of Word2 Vec word embedding technology to train the expanded Wikipedia corpus into the word vector,as the input of the follow-up model.(2)mapping of SQL query statements to dependency graphs: In this paper,the generation of SQL statements is converted to a filling problem in a dependency graph by using a dependency graph that is consistent with the SQL syntax,avoiding the order problem caused by the sequence-to-sequence formal model.(3)proposing a semantic driven NL2 SQL model: Based on SQL syntax dependency graph,Sequence-to-set and table list attention mechanism,this paper constructs the prediction model of SELECT clause and WHERE clause using LSTM neural network respectively.(4)intelligent recommendation for visualization of high dimensional data: This paper implements the automated recommendation process based on pivot table,table lens,table algebra and table name attention mechanism,which includes chart type,visual coding,visual attribute and visual mapping formula.In order to verify the validity of semantic driven large data query and intelligent visualization research,and because the WiKiSQL data set is extended in this paper,the SDM model is compared with the original model Seq2 SQL of WiKiSQL data set,the accuracy of generating SQL queries is comprehensively evaluated through two indicators of execution accuracy and logical form accuracy.Finally,an instance and NDCG indicators are used to evaluate the automatically generated visualized graphics.The experimental results show that the accuracy of the proposed SDM model is 15% higher than that of the Seq2 SQL model.At the same time,the recommended visual graph enables users to find answers to questions quickly,and to obtain information related to them without the need to ask questions again.