The upload screen for Quick Check within Westlaw Edge. Courtesy: Thomson Reuters. The upload screen for Quick Check within Westlaw Edge. Courtesy: Thomson Reuters.
|

Last year's release of Westlaw Edge from Thomson Reuters saw the widely-used legal research platform receive a number of tech-enabled upgrades, from new algorithms for search to tracking statutory changes and invalid law. But one component of Westlaw Edge was delayed for further refinement, and is now ready to be released: automated brief analyzer Quick Check.

Quick Check allows users to upload briefs, memos or motions into the Westlaw Edge system, where an artificial intelligence system will scan the document to detect the legal issues covered. Following a search that takes an average of one minute, Quick Check will then provide a report offering potential cases, briefs and other sources that may have been missed concerning the topic; point out cases cited in the uploaded document that have negative KeyCite warnings; and create a table of authorities. The recommendations the system provides include sources that are already found elsewhere within the document.

Thomson Reuters said Quick Check will be available as an additional free capability to Edge subscribers on July 24. The company has no plans at this time to offer Quick Check as a standalone product, or to make it available to users of other versions of Westlaw.

Carol Jo (CJ) Lechtenberg, director of Westlaw Product Management, told reporters that Thomson Reuters expects users to go to Quick Check for four main use cases. Naturally, there is the ability to update an old brief or memo, analyze an early draft, or make a final check before submission. Interestingly, though, she also expects the tool to be used to analyze opponents' work, essentially utilizing Quick Check to find where the opposite side in litigation have missed cases or may be using outdated law.

“This is something that customers have been incredibly excited about. … Any time they can take a look at their opponents' work and get a jump start on figuring out how to respond to a brief their opponent has submitted or figure out some holes in what an opponent has done is hugely beneficial,” Lechtenberg explained.

Once users have received their report output from Quick Check, the recommendations are broken down by headings that are extracted from within the document itself for easier search. There are also notes for the outcomes of a particular recommended case or particular issue within a case, which Lechtenberg estimated will come with about two-thirds of recommendations. With this, she said, “They can determine really quickly whether this particular recommendation is going to be favorable to them.”

There are then further options to filter the recommendations as desired, notably with eyeglass, folder and annotation icons that indicate how the user has interacted with the case in the past. Notes such as “frequently cited,” “last two years,” and “high court” also provide ways to filter content.

Mike Dahn, senior vice president of Westlaw Product Management, said development of Quick Check began in 2016 and was originally planned to roll out alongside Westlaw Edge's wider release. But often, testing found that the recommendations were providing source material that was already found by their researchers. The goal of the product, he said, was to be a supplement to the work legal researchers are already performing.

“They're experienced researchers, so they spend 5 or 10 hours doing that. … We worked hard to ensure that what we're showing them is highly relevant and goes beyond traditional research methods,” Dahn explained.

With that in mind, this is not a tool to begin research, but rather refine research once it's already underway. Lechtenberg noted that the system requires at least two citations in the uploaded document before the artificial intelligence can run.

Once there, though, the AI system responds with more than 10,000 grades by attorney-editors, KeyCite, and other sources of Thomson Reuters information to train the system. Tonya Custis, senior director for research at Thomson Reuters' Center for AI & Cognitive Computing, called it “by far the most data sources we've used on any AI project that I've been involved in.”

The AI system segments the document, then runs through four stages to produce the report: document analysis (segmenting the document), discovery (integrating external data), ranking (first finding 200 top cases, then more rigorous ranking), and recommendation (thresholding that grades the cases).

The end result, Thomson Reuters hopes, is a new offering that allows AI to work to empower legal researchers, rather than supplanting them. “Any attorney that you speak to will say that at some point in their career—some more than others—they've had these feelings at times that they're afraid that perhaps they've missed something in their research,” Lechtenberg said.