Longtime readers of SCOTUSblog are by now familiar with Casetext’s legal search tool. It solves an ever-present need for our team: finding opinions from all levels of the court system for our articles and case pages. Practitioners who read this blog, on the other hand, face a different need in their day-to-day work with the law. Rather than searching cases by name, attorneys need a way to search case law to find support for specific propositions. This task is challenging not just because the common law is vast, but because judges will use different articulations for the same proposition or principle.

Casetext addresses this formidable challenge head on with their new tool: Parallel Search. As opposed to simple keyword search, the limitations of which most of us are intimately familiar with, Parallel Search uses machine learning technology to match full phrases and sentences with those with similar meanings in case law, even if the results and query have almost no words in common.

It’s so powerful that users have described the technology as “straight up witchcraft” and “almost … like cheating” (though it certainly isn’t).

Need a federal court of appeals ruling for a protest case? Search for an argument like “The right to protest is fundamental to American democracy” and you’ll get jurisdiction-specific results you would never find via traditional search: opinions holding that “The right of citizens to petition their government for the redress of grievances is fundamental” or “A bedrock First Amendment principle is that citizens have a right to voice dissent from government policies.”

The secret is Casetext’s application of cutting-edge machine learning technology to the law. Transformers are artificial neural networks that “learn” by being trained on a pre-selected dataset. Parallel Search uses two layers of transformers that are each pre-trained on the entire body of U.S. judicial opinions. The result isn’t just accurate. It’s fast and, most importantly, affordable.

Introduced as part of Casetext’s brief-writing solution Compose, Parallel Search proved so popular with users that it is now available as a standalone solution. Interested? A free trial is available, as well as full pricing information and a demo for practicing attorneys.

Posted in Featured

Recommended Citation: Tom Goldstein, Legal research no longer limited to keywords, SCOTUSblog (Nov. 30, 2020, 12:37 PM), https://www.scotusblog.com/2020/11/legal-research-no-longer-limited-to-keywords/