Research on Tobacco Data Query Application Based on NL2SQL

Natural Language to SQL (NL2SQL) is an important direction in the field of natural language processing. This paper focuses on Chinese and tobacco industry data queries as research subjects and constructs a domain-specific natural language data query model. The Chinese NL2SQL task is deconstructed into two classification problems, using a general classification model and a conditional value retrieval model to combine and form a complete SQL statement. Additionally, BiLSTM and attention mechanisms are utilized to generate feature vectors, improving the accuracy of SQL statement generation. Moreover, a vector library containing tobacco industry vocabulary is trained, and a tobacco industry dataset is constructed. Experiments on this dataset show that the model proposed in this paper has higher accuracy and can effectively meet the data query needs of the tobacco industry.