Suggestive Coding System

The BIA Suggestive Coding System for Legal Issues and Topics

 

The BIA Suggestive ("Predictive") Coding engine allows your document review attorneys to manage workflow more effectively and prioritize highly-relevant data for a given matter through the marriage of trained legal judgment and algorithmic science.

 

The BIA Suggestive Coding System provides for the classification of documents based upon any criteria the user specifies.  The analytics engine can be seeded (although seeding to begin the process is not required) with any number of factors, together or separately:  responsive terms, hot documents-that is, those documents expressly known to be responsive or privileged or whichever criteria to which the user is looking.  BSCE then predicts the likelihood of a given coding for the remainder of the document collection based upon a subsequent sampling/QC process (e.g., via human reviewers).  It is a learning-aware tool so as more documents are provided to the engine, the accuracy levels increase.

 

The BIA Suggestive Coding System can provide a significant increase in the efficiency of the document review process itself and is not tied to any document review/hosting platform.  Any given document review can be organized in a manner that incorporates the results of BSCE into the set of documents to be reviewed.  Review rates can be increased by a factor two, three or more.

 

The BIA Suggestive Coding System is an implementation of Naïve Bayes classification that is enhanced by both algorithm specific modifications and industry specific modifications which are based upon its extensive knowledge in e-discovery and its extensive experience performing and managing large-scale document review projects, ranging from 300,000 to 4 million documents.   The system is also usable and effective in document collections as small as 5000 documents.

 

Real World Case Summaries:

 

Case 1: (Product liability matter with 30 custodians)

• Number of documents with eyes on review:  32,273 representing 17% of the total document universe.
• BIA Suggestive Coding was then applied to the remainder of the document collection of 160,727 documents that were previously reviewed.
• Overall accuracy of the process:  91.9%.  (Note the overall accuracy is determined through iteratively running the system during the review process and coding the 32,273 documents the system was most confident about.)  15 Primary terms were identified during the eyes on review and added to the algorithm.
• For those documents identified in the least likely to be responsive grouping yielded a 98.9% accuracy rate.

 

Case 2: (Intellectual property claim matter with 15 custodians including 5 attorneys)

• Number of documents with eyes on review:  30,000 representing 22.7% of the total document universe.
• BIA Suggestive Coding was then applied to the remainder of the document collection of 102,000 documents that were previously reviewed.
• Overall accuracy of the process:  93.4%.  (Note the overall accuracy is determined through iteratively running the system during the review process and coding the 30,000 documents about which the system was most confident) 32 Primary terms were identified during the eyes on review were added to the algorithm.
• For those documents identified in the least likely to be responsive grouping yielded a 94.0% accuracy rate.

 

Case 3: (Product liability matter with 202 custodians)

• Number of documents with eyes on review:  100,000 representing 29.3% of the total document universe.
• The BIA Suggestive Coding was then applied to the remainder of the document collection of 241,000 documents that were previously reviewed.
• Overall accuracy of the process:  87.4%.  (Note the overall accuracy is determined through iteratively running the system during the review process and coding the 100,000 documents about which the system was most confident.)  19 Primary terms were identified during the eyes on review were added to the algorithm.
• For those documents identified in the least likely to be responsive grouping yielded a 94.2% accuracy rate.

 

In the October, 2011 issue of Law Technology News magazine, New York-based U.S. Magistrate Judge Andrew Peck remarked, "Until there is a judicial opinion approving (or even critiquing) the use of predictive coding, counsel will just have to rely on this article as a sign of judicial approval. In my opinion, computer-assisted coding should be used in those cases where it will help 'secure the just, speedy, and inexpensive' (Fed. R. Civ. P. 1) determination of cases in our e-Discovery world."

 

To learn more about our products or to request a demo, please contact our Sales Department at 888.338.4242 or info@biaprotect.com.