BGL - Natural Language Processing for Financial Sector - SnT

This research explores the use of Natural Language Processing, AI, and Intelligent Conversation technologies to enhance automation in the financial sector.

The project at a glance

Start date:

01 Jan 2024
Duration in months:

55
Funding:

Industrial Partnership / FNR – Luxembourg
Principal Investigator(s):

Jacques KLEIN

About

Processing (NLP). In collaboration with BGL BNP Paribas, this research investigates financial document understanding and automatic question answering (QA) over challenging financial documents, including text, tables, and visual elements (e.g., graphs, images). By leveraging cutting-edge technologies such as Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG), we explore their application in the financial domain with a focus on reliability, trustworthiness, and strict adherence to compliance in this highly regulated sector. Build upon advancements in natural language processing (NLP) and large language models (LLMs), our research focuses on enhancing the automation of financial document processing. By integrating techniques such as Text-to-SQL translation, we aim to facilitate seamless interactions between natural language queries and financial databases, thereby bridging the gap between business needs and software capabilities. Additionally, we are developing efficient machine translation models tailored for low-resource languages like Luxembourgish, leveraging knowledge distillation to ensure performance within computational constraints. Our comprehensive approach also includes revisiting code similarity evaluation metrics, particularly focusing on Abstract Syntax Tree (AST) edit distance, to enhance the reliability and trustworthiness of automated financial systems. Furthermore, we address the challenges of function calling routing and invocation in LLMs by introducing a novel dataset and benchmarking state-of-the-art models, aiming to improve API function selection and parameter generation in complex, multi-step tasks. This multifaceted strategy underscores our commitment to deploying cutting-edge technologies in the financial domain, ensuring compliance and robustness in this highly regulated sector.