Data Query and Mining Technology based on Computer Natural Language Processing and DSL

This study proposes an integrated NLP and DSL framework to enhance data query and mining in the tobacco industry. Traditional query methods face efficiency and accuracy challenges when processing domain-specific natural language requests. The system combines: (1) A BERT-based NLP module for intent recognition and entity extraction using industry terminology; (2) A customizable DSL for structured query formulation; (3) An automated converter generating optimized SQL with context awareness; (4) Data mining employing Pearson correlation and regression for trend analysis. Experiments showed the framework reduces average query time by 72.7% (502ms vs 1836ms for manual SQL) while maintaining 96.5% semantic accuracy. The mining model achieved 96.3% prediction accuracy with 18.9% error reduction during training. The solution effectively bridges natural language interfaces with domain-specific data analysis, demonstrating significant performance improvements. Future work should expand testing to broader datasets to verify generalizability. This approach offers practical insights for developing specialized NLP-driven analytics systems.