The challenge
Federal rulemaking is a public process by design — agencies are required to accept and consider comments from anyone affected by a proposed rule. That openness is the point, but it creates a volume problem: reviewers can receive comments by the tens of thousands on a single rule, ranging from form-letter campaigns to detailed technical objections, all of which deserve fair consideration within the process.
Our approach
We applied AI/ML and natural language processing to the federal eRulemaking dataset to help reviewers triage and understand comment volume without losing the human judgment the process depends on. The goal was never to automate the decision — it was to surface structure in the data so reviewers could spend their attention where it mattered most.
- Natural language processing to group comments by theme and identify duplicate or form-letter submissions
- Machine learning models tuned to the specific vocabulary and structure of federal rulemaking language
- An interface built around the reviewer's existing workflow, not a new tool they had to work around
Why it matters
Public comment is one of the few direct channels between citizens and federal rulemaking. A platform that helps agencies process that input faster and more consistently protects the integrity of the process itself — it means every comment gets a fairer chance at being read and understood, not just logged.