Design and development of an LLM-based chatbot as a tool for outreach to the Congolese legal system

• Presented by Tshabu Ngandu Bernard
• Supervised by Professeur Antoine Bagula PhD
• Co-directed by Assistant Landry Mbale

Introduction

In an ever-changing world where technological advances are redefining learning and communication methods, the need to make the legal system accessible to all is becoming crucial.

However, access to legal information in the DRC is exacerbated not only by the scarcity of digital resources, but also by the intrinsic complexity of the law, which makes it difficult for the uninitiated to understand.

This situation is all the more critical given that legal information, essential to the exercise of rights and citizen participation, often remains confined to legal texts that are difficult to access and read. Understanding the law, with its technical terms and abstract concepts, requires pedagogical mediation to make it accessible to the general public.

Hypothesis

The use of an LLM-based Chatbot presents itself as an innovative solution. Thanks to artificial intelligence and its ability to interact conversationally, a Chatbot can offer clear explanations, answers to common questions and legal information in an accessible, user-friendly way.

Research questions

1. How can we fine-tune LLMs so that they faithfully reflect the linguistic nuances and cultural specificities of the legal framework in the DRC, ensuring maximum relevance and effectiveness in the local context?

2. How do you design a chatbot capable of simplifying and translating complex legal concepts accurately while making them accessible to a non- specialist audience, without compromising the accuracy and fidelity of the information provided?

3. What methodologies should we put in place to rigorously assess the reliability and accuracy of the data provided by the chatbot?