Migrated from eDJGroupInc.com. Author: Amber Scorah. Published: 2012-03-13 05:00:11Format, images and links may no longer function correctly. The paradigm of review is changing. Predictive coding has made a huge impact in the e-discovery and document review world. In this interview, I speak with Jeff Nadalo, Litigation Counsel at TransCanada Pipeline. He helps cut through all the marketing jargon to give an understanding of the technical and realistic implications of predictive coding, and how it can affect your e-discovery practice.
Amber Scorah: What do the courts think? Is predictive coding defensible?
Jeff Nadalo: I think we certainly are seeing that. There are some recent cases coming out which are indicating that predictive coding is certainly an acceptable means of minimizing the resources that it takes to perform large-scale discovery.
Amber Scorah: If your competitors are leveraging this technology, should you?
Jeff Nadalo: TransCanada has invested heavily in predictive coding and I think the key reason is that it can pay for itself in a single use on a large matter.
Take, for example, a case where you have millions of documents. In our experience, predictive coding can reduce that number substantially. We are talking about a million documents, after you have already gone through them and gotten rid of documents which are not relevant, just simply based on traditional discovery practices.
If you can take those core documents and eliminate having to have an experienced attorney or even a paralegal review those documents, the cost savings can be substantial.
Amber Scorah: When is predictive coding most effective?
Jeff Nadalo: The first is in large-scale discovery cases where you have multiple custodians over various departments, with multiple different issues. In those instances you can find yourself struggling simply to manage the files themselves.
I have worked on cases where we had to have more than 50 reviewers on files. When you can reduce that number from 50 to 5 highly trained attorneys, who can then really commit themselves to focusing on documents which are the most critical key documents, two things happen:
- you save on the cost of having to hire contract attorneys or paralegals to help you review what in most cases are 95% irrelevant documents; but more importantly,
- the attorneys that are working on the case can start focusing more effectively on the case and the merits of the case and really get into the documents and begin digesting and trying to understand those documents much better.
Amber Scorah: What are the various technologies that you have used and seen and what are your opinions on them?
Jeff Nadalo: We evaluated roughly a dozen technologies that are out there. We found that predictive coding can vary from one quite simple product, to something a little bit more complicated than a sophisticated search that adds some basic functionality, all the way up to coding software. This coding software is the type we use internally – it is highly sophisticated, using a complex algorithm that essentially trains itself.
When using these technologies, you can find yourself surprised by the results. Let me give you an example.
During our evaluation, I thought it would be interesting to use my own personal data. We ended up pulling in my own data and ran predictive coding on my work emails and also some personal emails that were in there. Some of the results were outstanding, and quite fascinating.
The software is capable of understanding relationships between documents that are frankly not apparent to anybody else. It is relying entirely on the way words are placed, how words interrelate to each other, and the types of words.
So it is not just a keyword search, and that is what is important to understand about predictive coding. The more sophisticated tools use an algorithm which is based on the characters, as well as where they are placed, how they are used, and the contexts that are used.
During our experiment with my data, I have found relationships which were quite accurate, and they reflected a relationship within the documents that was critical. And so had we decided to code them according to the suggestions that the software was providing, it would have been pretty much spot on. In fact, it was that demonstration of the capabilities which ultimately personally led me to select this particular vendor that we ended up using.
For more information or to register, visit www.ediscoveryoilandgas.com or email amber.scorah@iqpc.com.
Amber Scorah – eDiscoveryJournal Contributor