Manage Relevance setup in eDiscovery (Premium) (classic)

eDiscovery (Premium) Relevance technology employs expert-guided software for scoring files by their relevance. eDiscovery (Premium) Relevance can be used for Early Case Assessment (ECA), culling, and file sample review.

eDiscovery (Premium) includes components for the Relevance training and tagging of files relevant to a case. eDiscovery (Premium) learns from the trained samples of Relevant and Not Relevant files to provide Relevance scores for each file, and generates analytical results that can be used during and after the file review process.

Guidelines for setting up Relevance training

In Advance eDiscovery, in the Cases window, select a case and select Go to case. Select Relevance > Relevance setup. Follow these recommended guidelines to set up Relevance.

• Tagging: The effectiveness of the iterative Relevance training process is dependent on the ability of the expert to tag the file samples with precision and consistency.

• Case issues:

  • For each issue, use the same expert throughout the entire Relevance training process. Simultaneous tagging of the same issue by multiple experts isn't permitted.
  • Determine if each group of files is pertinent only to a specific issue.
  • If an issue is defined too generally, eDiscovery (Premium) may yield too many files that aren't relevant. If an issue is defined too narrowly, the Relevance training process may take more time.
  • During each Relevance training cycle, eDiscovery (Premium) focuses on a single active issue and interim sample results are displayed accordingly.
  • In a multiple-issue scenario, the Sampling mode enables the selection of issues to be included in processing. Issues defined as "off" aren't handled until their Sampling mode is changed. An issue can be "idle" or "on" for only one expert.
  • eDiscovery (Premium) can be used to generate candidate privilege files. Set up a separate issue for privilege. If possible, train and cull for relevance first, and then train for privilege on the culled set only (reload the culled set as a separate case).
  • Batch calculation can be performed only when there are no open samples (when selecting Batch Calculation, there will be a list displayed of users with open samples). To "close" samples of other users (this should be performed only if these users aren't tagging these samples), an Administrator can use the "Modify relevance" utility with the "All users sample" option.

• Metadata: eDiscovery (Premium) focuses on content. It doesn't consider metadata as part of the relevance criteria.

• Richness: If the Richness for an issue is less than 3% after Assessment, consider seeding the Relevance training with known Relevant and Not Relevant files.

• File size: Large files (over 5,242,880 characters of extracted text) are ignored in Relevance. The files don't participate in the Relevance training process and don't receive a Relevance score after Batch Calculation. Files over 5 MB can be included in the Assessment set.

Setting up case issues

The parameters described in this section are available in the eDiscovery (Premium) Relevance > Relevance setup.

• Issues must be assigned to a user who will train the files.
• Imported files must then be added to the load being processed.
• Define and organize issues carefully, as this can impact the Relevance training results.

After parameters are set, the reviewer / expert can start training the files in the Relevance tab.