Research And Implementation Of Algorithms And System On Automatic TXT2SQL

"With the development of the Big Data technology and Mobile Internet,Data has become a valuable asset of enterprises and individuals.Mining and analyzing date has become a centre stage in the Big Data Era.However,most data analysis softwares still require professional programming languages(SQL)to interact with the database,which blocks the development of data analysis technology.How to reduce the application cost has become a fascinating research topic in the industry and academic.Automatic TXT2SQL technology emerges as the times require,which provides user with intelligent interface function by transforming natural language query into executable query.However,there are still some challenges existing in the application of real scene:1)In the form of SQL,current methods only consider the simple pattern,but fail to effectively generate complex sentences.2)In the structure of database schema,the existing methods mainly focus on single table or same domain database,which can not be effectively applied to complex structure schema.3)There are few open APIs and systems for TXT2SQL,and the practice of model landing is still lacking.To solve the above problem,this paper focus on the research and implemenetation of algorithms and system on automic TXT2SQL.The main research works including:(1)Proposed a solution for complex TXT2SQL generation,including an intermediate representation network and a set of concise semantic query language,which solve the problem of complex pattern SQL query generation.In the public dataset,our solution outperform the state-of-the-art model by 19.5%.(2)Proposed a schema representation solution for complex database schema structure,including a type linking method and a graph-structure representation,which alleviates the generation problem in the complex structure schema.Our solution achieve 6.2%improvement on the public dataset.(3)Proposed a pre-consultation system of the medical ophthalmology based on artificial intelligence algorithm which designed and implemented in this paper.The system provides user with condition collection,diseases prediction and analysis,etc.Our system releases the precious time wasted on user consultation,and verify the two algorithms preprosed in a practical view.