Of Predictive Coding and Patents

Posted by Bob Tennant on Thursday, June 16, 2011

Recommind caused quite a stir last week when we announced that we had received a patent on our “systems and methods for predictive coding. ” Our competitors wasted no time raising objections. If nothing else, the ruckus acknowledged the importance of automation to the future of legal discovery, and the significance of our intellectual property, including the announced patent, to the eDiscovery industry.

Given that iterative, computer-expedited review is generally perceived to be the future of document review, the stakes are indeed high, which makes the competitive response logical – if somewhat disingenuous. These responses have been based on false premises and have served to obfuscate both the facts and our motivations. I would like to clarify the situation by addressing a few of the more common misconceptions here.

Some commentators claimed our patent might be difficult to defend due to the existence of prior art. In one blog post, ironically, the author cited a technique called LSI and work by Dr. Thorsten Joachims around support vector machines (SVMs). The author was apparently unaware that Recommind invented and holds another patent on PLSA and its uses—an algorithm that was invented to overcome some of the serious limitations of LSI—and that Dr. Joachims was an early advisor and technical contributor to Recommind. Another post cited work in relevance feedback, which is an ad hoc way of refining search results based on user clicks or feedback in search systems. While refining displayed search results based on apparent areas of user interest is a useful application in e-commerce and other kinds of search, it is not particularly relevant to the document review process, and more specifically to the methods we have patented, where documents are being worked through systematically. In fact, none of the algorithms described above directly address the process of conducting legal review.

Even more ironically, there are patents held on all of the techniques mentioned above and their uses that went unmentioned by these competitors. While one would presume that all of the competitors working in eDiscovery would have received appropriate licensing for techniques they may be using or claiming to use, unfortunately, we know for a fact that that is not always the case. Off the top of my head I can think of three competitors who have claimed in the past to use PLSA as the engine powering their search, which as I mentioned above, is an approach to searching owned by Recommind, and which we had not licensed to the companies in question. In each case we were able to resolve the situation without resorting to lawsuits, a pattern of behavior we prefer to maintain if we can.

Some have also attempted to rewrite the history of these technologies and their application to the discovery process. Recommind’s inventions in search and text classification, as mentioned above, date from the early part of the last decade. Our first patent filing was in early 2001, and the research itself dates from the late nineties—before most of the companies complaining today were even founded. The “systems and methods for predictive coding” patent is about applying these technologies in a process, one that we have developed over years of deploying our technology in practical circumstances and instantiated in software—but it is not the only process that can be utilized to improve document review. Other companies used some sophisticated computer techniques in the eDiscovery process before Recommind—in particular Dolphin Search and H5—but those techniques were not used in the way described in our patent. Dolphin Search is no longer around, but H5 is an example of an organization that we do not believe violates our recent patent, in no small part because the process, techniques and technology at the heart of our system provide a different approach to the batch process employed by H5.

Recommind has invested countless hours, millions of dollars, and a great deal of energy creating a system and method for iterative computer-assisted document review, so that review can be done in a predictable and defensible manner that saves its users a great deal of time and money and ensures less risk as a result of more accurate review. We have spent a great deal more energy evangelizing this approach to doing review. The USPTO’s grant of a patent on these methods in April rests on a foundation a decade in the making and affords us the protection of law in the defense of our property rights—the same rights that are afforded every organization in the United States and that are claimed in some form or other by all of the same competitors so upset by this patent. These rights protect our methods for improving the discovery process, not those of others. Once the dust settles and the industry has digested the news, it will become clear whether our methods are as valuable as we, and apparently our competitors, think they are.

For educational resources on Predictive Coding, please visit www.predictivecoding.com.