Legiscan
Stanford Law School / CODEX LLM hackathon project. Compares proposed legislation against status quo using GPT-3.5.
Overview
Software that implements the GPT-3.5 LLM to compare proposed legislation against the status quo in order to return a summary, concise overview of the net-change, as well as a complexity index score to the user.
The Challenge
Legislative bills are written in complex legal language and can be hundreds of pages long. Citizens and policymakers struggle to understand what bills actually do and how they differ from existing laws. Comparing proposed legislation to current law is time-consuming and requires legal expertise.
The Solution
Developed an NLP system using GPT-3.5 to analyze legislative text and generate plain-language summaries. Implemented comparison features that highlight key differences between proposed legislation and status quo. Created a complexity index that scores bills based on readability, legal terminology density, and structural complexity. Built for the Stanford Law School / CODEX LLM hackathon.
Impact & Results
Made legislation more accessible to general public
Automated comparison between proposed and existing law
Provided objective complexity metrics for legislative review
Generated concise net-change summaries
Developed at Stanford Law School / CODEX LLM hackathon