With the increasing amount of data being collected and stored by organizations, the challenge of finding, retrieving, and transforming that data into useful insights has become more complex. Many companies struggle to fully utilize their data because employees face obstacles like not being able to find what they need or translate their questions into the code needed to access the information.
To address this issue, generative AI technology is being developed to simplify the process. Large language models (LLMs) are being trained to write SQL, the dominant language for interacting with databases, making it easier for more users to access and analyze data. IBM has made significant progress in this area, with their code model ranking at the top of benchmarks that measure the AI’s ability to translate natural language questions into SQL queries to run on real data.
While IBM’s text-to-SQL generator is not yet as accurate as human-generated SQL, it shows promise and has the potential to narrow the gap as AI continues to improve. The system developed by IBM researchers uses a three-step process to extract schema information from databases, generate SQL code, and execute queries, providing users with insights faster and more efficiently.
In addition to the text-to-SQL generator, IBM is also working on developing a conversational graphical user interface (GUI) that allows users to interact with their data more intuitively. By combining the personal touch of an AI chat interface with the simplicity of a web-based GUI, IBM aims to make it easier for data engineers, managers, and analysts to explore and analyze structured data.
The conversational GUI features a chat interface on the left that aligns with a graphical representation of the data on the right, allowing users to ask questions and review results seamlessly. The system provides users with suggested tables, sample questions, and visualizations to help them refine their queries and understand the insights generated by the AI.
Overall, IBM researchers are working on incorporating these generative AI capabilities into their Watsonx products and are continuously improving the text-to-SQL generator and underlying language models. By leveraging AI technology to simplify data retrieval and analysis, IBM aims to help organizations unlock the full value of their data and make better business decisions.
Article Source
https://research.ibm.com/blog/granite-LLM-text-to-SQL
