Legal Analytics, The Next Frontier: Using Language Analytics to be More Persuasive in Court

Legal analytics—the use of data to make quantitative legal predictions that inform decisions—is currently moving from the margins of the profession to the mainstream. Underlying the rise of analytics is the maturation of artificial intelligence technologies like natural language processing and machine learning, which are deployed to add structure to complex legal data, which in turn can be used for statistical analysis.

At its core, most legal data is relatively “unstructured,” lacking tags or metadata that help computers understand meaning in documents. Now, with advances in machine learning, attorneys, editors and other subject matter experts can help train computers to add the missing detail to structure vast amounts of legal data, enabling machines to replicate human editorial activities at scale. With clean, structured data, companies can then create powerful new tools that identify important legal trends and help attorneys make better legal and business decisions.

The use cases for legal analytics are growing. While mining dockets allows attorneys to do a better job selecting venues, evaluating opposing counsel and making facts-based business decisions, new analytics based on extracted data from case law and the opinions written by judges can help attorneys prepare better arguments and smarter legal strategies for litigation. Whether attorneys are looking for language that certain judges have found persuasive in previous cases or looking for the best way to challenge opposing counsel's expert testimony, they can use legal analytics to identify trends and patterns in the language used by attorneys and judges—and associated outcomes—to improve their practice of law.

In Analytics, Comprehensiveness and Quality Matter

Adding structure to data is a difficult and expensive process. A docket or a case opinion often contains a lot of rich—albeit unstructured—data which was written for consumption by humans, not machines. Natural language-based data is, of course, searchable by computers, but computers lack contextual understanding.

When you simply attempt to “string match”—e.g. search for keywords—you often get a mix of results, both relevant and irrelevant, and that means you have to spend considerable time sifting through the initial results set. For example, when a researcher wants to find only those cases in which a motion for summary judgment is granted, turning up every case that includes both the words “motion for summary judgment” and “granted” is simply over-inclusive and will lead to lots of incorrect or irrelevant results. However, if you can train algorithms to look for specific indicia, or use natural language processing to decompose syntax and find only those rulings where a motion was adjudicated, suddenly you have a very precise analytical toolset that can isolate and aggregate only the most relevant data, generating accurate insights into the ways judges rule or law firms perform.

The training process takes a huge commitment and significant investment. Accuracy in machine learning is both a reflection of the quality and comprehensiveness of the data inputs on the one hand, and the quality of the training data on the other. As the saying goes: good data in, good data out. Legal data is, unfortunately, often filled with errors. Apart from the errors, it is also very complex.

To both clean the data and structure it, companies have to invest in large data collection, clean-up and editorial processes. The editors who do the human part of this work must have expertise in the activities they are performing—either knowledge of specific areas of law or linguistics—to add the appropriate structure to the data. When you combine that human expertise with comprehensive and reliable data, advanced analytical technologies, expert-created classification systems and expert-created training data, then you can begin to mine the data for reliable insights that were often simply inaccessible before.

Seeing the Patterns in the Ways Judges Rule

Understanding how judges rule has long been a key to having a competitive edge in litigation. Partners at firms that litigate repeatedly in certain venues aggressively market their superior knowledge of specific judges, an advantage that gives them the best chance at winning. Similarly, law firms may pay a premium for attorneys who recently clerked for one or multiple years writing and researching for a specific judge, in part because of the knowledge those clerks gained in being the “voice” of the judge in question. But, when spread across courts, jurisdictions, and years, this anecdotal knowledge does not scale, and presents a challenge for even the most profitable of firms.

Firms are now employing advanced analytical tools that can provide similar data on a comprehensive scale. After all, there are patterns in the way judges write and rule. Knowing those patterns can be the difference in winning or losing a case, whether it is the cases a judge cites most frequently, the frequency with which they grant a certain motion type, or the actual language they cite to time and time again. One senior litigator at a prestigious litigation boutique firm noted that language analytics taught him that a certain judge dislikes sports analogies; that attorney now refrains from saying things like “moving the goal posts” when arguing before that judge.

Language-based analytics can also help change judges' opinions. An East Coast firm recently was arguing a case in California, where they don't have an office. They submitted an argument that was quite common in New York, where they are based, but the California judge summarily dismissed the argument as “atypical” in his court. Using analytical tools that show the history of the judge's rulings, the firm was able to find several prior occasions where the very same judge had not only allowed, but ruled in favor of, a similar argument. When attorneys for the firm produced the evidence, the judge begrudgingly accepted and reversed the ruling. The firm believes this was ultimately a critical decision that helped their case.

Law firm partners are also using legal analytics tools to better inform their clients. A senior partner at one firm routinely looks up judges while on the phone with clients, informing the client of the likelihood of success of certain motions, guiding the firm's litigation efforts by directing attorneys to file only those arguments with high chances of success. This saves clients money, and don't think clients haven't noticed. Litigation spending remains the hardest legal cost to forecast, and corporate counsel, often the largest clients of large law firms, are placing a premium on firms that use analytical tools to help keep costs down and increase transparency in their litigation strategy.

Part 2 of this article will examine additional ways legal teams are using analytics to identify trends and patterns to improve their practice of law.

Nik Reed is the co-founder and COO of Ravel Law, a legal research and analytics platform acquired by LexisNexis in 2017. While at Stanford, he co-founded Ravel, as a joint research project between the law school, the design school, and the computer science department, with the goal of using modern AI based tools to help attorneys better perform legal research. Prior to Ravel, Nik worked in management consulting where he focused on international mergers and acquisitions in the technology industry.