Scholarlink

As databases became more complex and widely utilized, the demand for user-friendly systems that allow non-technical users to engage with data is increasing. Current Text-to-SQL solutions frequently struggle with complex operations, especially those that necessitate the capabilities and versatility of PL/SQL, restricting their usefulness in real-world situations characterized by elaborate queries. To fill this gap, we suggest a dual-model architecture based on a transformer-driven sequence-to-sequence (Seq2Seq) framework. Our system effectively assesses the complexity of a specific query and chooses either SQL or PL/SQL for execution, improving its flexibility and user-friendliness. Utilizing the T5-small model, the approach attains 72% accuracy for PL/SQL queries and 90% for SQL queries, showcasing a notable enhancement in managing various database operations. This study addresses the shortcomings of previous research, offering a strong and comprehensive method for database interaction, which enhances the versatility and practicality of Text-to-SQL systems. Our efforts not only overcome the shortcomings of current techniques but also establish a basis for developing sophisticated systems that can handle various database management duties.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Leveraging Transformer Models for Generating PL/SQL Queries

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Namita Bhat,
  • Vaishak Kolhar,
  • Abhay Hosakoti,
  • Shreya Das,
  • Uday Kulkarni,
  • Shanshank Hegde

摘要

As databases became more complex and widely utilized, the demand for user-friendly systems that allow non-technical users to engage with data is increasing. Current Text-to-SQL solutions frequently struggle with complex operations, especially those that necessitate the capabilities and versatility of PL/SQL, restricting their usefulness in real-world situations characterized by elaborate queries. To fill this gap, we suggest a dual-model architecture based on a transformer-driven sequence-to-sequence (Seq2Seq) framework. Our system effectively assesses the complexity of a specific query and chooses either SQL or PL/SQL for execution, improving its flexibility and user-friendliness. Utilizing the T5-small model, the approach attains 72% accuracy for PL/SQL queries and 90% for SQL queries, showcasing a notable enhancement in managing various database operations. This study addresses the shortcomings of previous research, offering a strong and comprehensive method for database interaction, which enhances the versatility and practicality of Text-to-SQL systems. Our efforts not only overcome the shortcomings of current techniques but also establish a basis for developing sophisticated systems that can handle various database management duties.