The Atlanta Lawyer August/September 2019 - Page 15

Georgia State University College of Law. “The grandchildren write about what’s new in their lives, about their schoolwork, their hobbies, and where they spent their most recent vacation. While she is interested in everything her grandchildren do, she needs only certain pieces of information on which to base her own actions: whom to send money, whom to ask for pictures, and so on.” "Analytical tools could help her to organize the letters and conceptionally cluster the information contained in them," says Smelcer. Similarly, analyzing hundreds (or more) legal documents “can reveal patterns not apparent for the human brain – connections between words, combinations of words and phrases, and references - that help to extract rules from the unstructured, inherently messy legal language,” states Smelcer. Wait a minute, I hear you cry, how can the exquisite writings originating in well-educated lawyers’ minds, their skillfully built argument structures peppered with wit, wisdom, and profound considerations, be reduced to a data cluster or even a mathematical formula? “The machines do not look at an individual document,” explains Ben Chapman, Executive Director of the Legal Analytics & Innovation Initiative at the Georgia State University College of Law. “Rather, they detect connections between many documents, similarities, and relationships that mean something. They do not spit out a judicial opinion,” he clarifies, "they create models that help predict a certain future outcome." Why do we need Legal Analytics? Such quantitative predictions have value, e.g., for a law firm’s budget and staffing decisions and for calculating the risks of a matter. Legal Analytics is mostly Litigation Analytics. It is used to assess the time frame, potential outcome, and costs of a lawsuit; to predict the judges’ behavior based on precedent; and to evaluate and select credible expert witnesses based on depositions, trial transcripts, as well as jury verdicts and settlements. In addition, it is also used by law firms to analyze their competitors and adjust their resources and strategies accordingly. Legal Document Data Lawyer Interprets Data Strategic Decisions Business Development/ Client Management What can Legal Analytics do and NOT do? Machine Learning classifies decisions as raising a particular legal issue and helps with retrieving similar cases. The Lex Machina program that was developed at Stanford University originally predicted outcomes of Intellectual Property claims based on a corpus of all IP lawsuits in a ten-year-plus period. It then analyzes certain features of the cases such as the identity and behavior of the participants of the suit. The program also reads and organizes data to help users gain “insights and strategic advantage” in federal antitrust litigation, comments Rachel Bailey, Legal Data Expert at Lex Machina. Ravel, another Legal Analytics program, creates visual maps of Litigation cases, so-called network diagrams. Citation networks include the judicial history – which cases or arguments did the judge find most persuasive, what were the rulings, and what specific language did she use. Statutory networks detect relations among entities referred to by or subject to a particular regulation across multiples statutes and jurisdictions. Social networks examine communication relations among entities, e.g., senders and receivers of corporate emails, which is predominantly used in E-Discovery. Legal QA: Modeled after IBM’s “Watson,” the Ross program searches large text collections to locate “snippets,” i.e., documents, short phrases, or sentences that directly answer a user’s question. Like Watson, Ross learns from user feedback. (Continues on page 17) THE ATLANTA LAWYER 15