Stringent quality control with unwavering alignment.
Relativity structural and conceptual analytics, including Relativity assisted review (predictive coding), play a pivotal role in the development of the Strategic eDiscovery intelligent review workflow. Combining the consultative expertise of the Strategic eDiscovery support team and Relativity technology results in a superior review workflow customization unparalleled in the industry.
Relativity Structural Analytics
Relativity structural analytics better prepare your documents for review.
- Email threading organizes emails that were part of a single conversation and identifies the inclusive documents within a thread. Batching only the inclusive or end e-mail in preparation for the 1st pass review will reduce the number of documents for review by 25% or more.
- Near-duplicate detection can assist the reviewer with quickly finding documents with highly similar content.
- Foreign language identification can batch documents by language or in preparation for predictive coding.
Relativity Conceptual Analytics
Relativity conceptual analytics improve review efficiency and enhance quality control.
- Categorization allows documents to be sorted into conceptually related batches based on user selected examples of those categories.
- Clustering provides a high-level overview of the topics found in each matter.
- Concept searching allows the reviewer to submit relevant documents or passages of text to the analytics engine to identify conceptually similar documents within the collection, thereby significantly reducing review time. Documents can be sorted into related batches without user input.
- Repeat content identification items such as e-mail footers can be automatically identified and filtered out of the data set.
- Predictive coding utilizes Relativity Assisted Review, an analytics component of Relativity. Predictive coding prioritizes documents so the most relevant are reviewed earlier; it also typically reduces document volume.
Relativity Assisted Review: Predictive Coding
There are several parameters for determining whether a project is suitable for predictive coding including the type of matter, its risk profile, the availability of outside counsel and the timeline and the document volume and richness. We have significant expertise in the evaluation of the suitability of predictive coding for an assignment and the implementation of the processes thereof, if appropriate.
Our predictive coding processes and workflows are collaborative and transparent, and we provide extensive reporting throughout to ensure that legal analysis and judgment is based upon all available information, and that the entire process is defensible.
Predictive Coding Workflow
- Attorneys with substantive knowledge of the matter create a “test” set of documents, used to inform the predictive coding engine on the baseline of responsiveness.
- The attorneys code “seed sets” on an iterative basis to further teach the predictive coding engine the meaning of responsiveness.
- The iterative process is repeated until “stability” is achieved; subsequently, the decisions incorporated in the seed set are technically replicated through the balance of the document population.
The output of the predictive coding engine is the ordering of documents from most responsive to least responsive. Then the project team conducts issue coding, beginning with the documents receiving the highest responsiveness score. At several points, the team identifies the richness (percentage of responsive documents) and precision (percentage of correctly coded documents) to determine when to cease the full review and begin sampling of the remaining documents.