Application of LSTM-Based Seq2Seq Models in Natural Language to SQL Conversion in Financial Domain

Abstract

As a crucial branch of artificial intelligence, Natural Language Processing (NLP) enables computers to understand, interpret, and generate human language, significantly enhancing the efficiency of information retrieval and search. Given the growing demand for data processing in the financial sector, this paper proposes and implements a Seq2Seq model based on the LSTM algorithm to convert natural language queries into SQL statements (NL2SQL) for application in finance. The model demonstrates stable and significant performance improvements over 10 training epochs, with accuracy increasing from 0.75 to 0.9877 and the loss value decreasing from 1.5 to 0.4978. These results validate the accuracy and effectiveness of the proposed LSTM-based Seq2Seq model in handling NLP tasks within the financial domain.