Exploring Government Regulations with RAG LLM

Challenge

Many sectors of the U.S. economy are overseen by a morass of government regulation scattered across different laws, policies, memos, and handbooks. Navigating these regulations can be a time-consuming and challenging task. Nowhere is this issue more pronounced than in the housing sector, especially in the regulation of affordable rental housing. The complexity is staggering - just one section of a single HUD housing handbook spans almost 800 pages. Even for the government employees tasked with interpreting and applying these rules, the challenge of efficiently navigating this regulatory maze is daunting.

‍

Recognizing the potential of Large Language Models (LLMs) to address this issue, Element14 sought to explore their application to synthesize and navigate housing regulations. However, given the vast scope of regulations, they needed a way to test the feasibility without committing to a large-scale project.

‍

‍

Impact

To explore the viability of this technology, we collaborated with Element14 on a rapid prototype. Our focused approach resulted in a working prototype within just six weeks. We identified a subset of seven individual regulations governing Section 8 Housing for the prototype and built a Retrieval-Augmented Generation (RAG) model. We wireframed and developed a lightweight front end using our existing Azure infrastructure.

‍

This prototype served as a proof of concept, demonstrating the potential of LLM technology in a quick and lightweight manner. The resulting tool provides a comprehensive, user-friendly solution for accessing complex regulatory information. This prototype showcases the power of LLMs in simplifying regulatory complexity and opens up possibilities for similar applications across other government sectors.