Followers of H5 know the value we place on expertise—especially as it relates to keyword search. We say it loudly and often—not as a self-serving exercise, but because we know it to be true: successful keyword search for e-discovery requires linguistic expertise.
Litigators who aren’t convinced should carefully consider Kimberly Johnson’s insightful October 30 ACED’s piece, “This isn’t Google: E-discovery optimizers, keywords, and why you need a linguist,” in which she opin
es on the dangers faced by attorneys who may overestimate their facility with keyword searches. In fact, one could infer from her article that the “ethical duty of competence” to which lawyers must adhere may actually rest in acknowledging that their own language skills and search proficiencies, no matter how stellar, don’t rise to the level of expertise required to do a good job of searching when it comes to e-discovery:
“A lawyer can no more infer a mastery of e-discovery keywords from her aptitude for querying Westlaw than a whaler can presume that his skill set qualifies him to trawl for shrimp — harpoons are not nets. To rephrase in the parlance of linguistics, precision is not recall, and the search for relevant documents turns upon recall. An attorney’s above-average feel for language might find a document in a database, but the process of capturing reasonably complete sets of relevant documents within a large corpus entails a different set of skills.”
Considering the goals of search in e-discovery, the distinction Johnson makes between precision and recall is a vital one. If you think you’ve competently conducted a search because nearly every document hit is deemed relevant (high precision), think again. High precision probably means you’ve left quite a bit behind (poor recall) because precision and recall tend to be inversely proportional. And although the search for relevant documents may “turn upon recall” as Johnson suggests, so does the risk of cost overruns for reviewing the non-relevant material that will be caught in the net. Adding linguistic expertise and an iterative process is your best bet for achieving both a high precision and high recall result, one that will ensure that obligations are met while costs are contained. If you’re still not convinced, we again direct you to the TREC Legal Track studies.
Where linguists really shine in e-discovery, though, is where they’re needed most: finding the evidence. Addressing recall and precision in document review is one thing, scouring a document population for the evidence that will prove or disprove the fraud, the stolen IP, the malicious malpractice, the insider trading, is quite another. Like psychologists digging into the psyche, linguists know how to peel through the layers of vocabulary, syntax and lexical configuration that usually give it all away. “Hot docs” are more effectively targeted than stumbled upon, and the linguist can both help define the target and deploy the appropriate weapons from the linguistic arsenal to hit it. As an example, Johnson points to an H5 project in which our linguists successfully targeted verbs rather than nouns to yield results in a fraud case. Something counsel would have figured out? Unlikely.
By now, there are many cases where search expertise in discovery may very well have helped avoid a loss and/or sanctions. (See, for example, Victor Stanley v. Creative Pipe, In re Fannie Mae, In re Seroquel, Kipperman v Onex Corp). Johnson’s article is a good reminder that even when counsel has ample know-how and good intentions, there is no substitute for expertise. As the court noted in Kipperman v. Onex as it denied relief to the defense from having to review and produce the voluminous results–much of it irrelevant–from an “inartful” search: “Defendants must now lie in the bed that they have made.”
To up your chances of a good night’s sleep, call in the experts.