A Natural Language Interface for making Investment Decisions with Business Intelligence Chart and SQL Generation using LLMs
The Challenge
A investment bank aims to develop a Question Answering system capable of responding to natural language queries about stocks. Tatras utilized an open-source LLMs designed to transform natural language questions into valid SQL queries. This system retrieved the necessary data and presented the results in natural language, along with generating a graphical representation of the data. The primary challenge lies in accurately fetching results from the company’s database while minimizing the rate of incorrect information.
Hypothesis
- An open-source LLMs can effectively take a natural language query and generate a SQL query that will run against the database.
- Another LLM can assist the SQL generation llm by creating a chain of thought and incorporating relevant examples into the prompt, thereby improving the accuracy and relevance of the SQL queries generated.
Execution
- Ensure the LLM generates accurate SQL queries.
- Enhance the LLM’s ability to understand historical queries.
- Ensure that the generated charts, COT response are accurate.
- Implement question categorization to improve LLM performance.
- Libraries used: Langchain, PyTorch, Transformers, vector databases, SQL Databases and Databricks.
Outcomes
- A chatbot that processes natural language queries and provides its answers.
- Also provides charts, and chains of thought that enables users to visualize data for better comprehension of chatbot responses.
- The chatbot is able to retain context from the previous conversation history.
- A GenAI-based intelligent chatbot capable of understanding complex user queries and generating responses.
- Achieved significant cost savings by utilizing open-source LLMs.
- Ensured data privacy by keeping all data within the premises.
Project Highlights
65%
increase in predictive accuracy over existing human based approach.