Environment and workspace setup

Before creating an Active Learning project, you must first verify that your system and workspace meet the necessary standards, and then perform the required installation and configuration steps to successfully run an Active Learning project.

This page contains the following information:

Installing the Active Learning application

Install the Active Learning application from the Application Library to your workspace. For more information, see Installing applications from the Application Library.

Agent configuration

Ensure the following agents are installed and configured:

For more information, see Agents.

Relevant instance setting table values

Active Learning uses the following instance settings:

  • ReviewQueueRefreshThreshold
  • ReviewQueueBatchSize
  • ClassificationCategorizationDelay
  • ClassificationCategorizationMaxDelay

Required workspace components

A new Active Learning project uses the following components, so you must create them before you can create a project. Even if these items already exist in the workspace, you may want to create a new instance of each specifically for your new project.

Saved search

This saved search includes the documents to be used in the Active Learning project. This set of documents is used as the searchable set when you create an Analytics index. The documents must contain extracted text, and the searchable set must contain example documents to train the model. For more information, see Pre-project sampling.

Note: The saved search must be public.

Analytics index

You must create an Analytics index with Classification as the Index Type. You must create a separate Analytics index for each Active Learning project. For more information, see Creating an Analytics index.

Note: The Analytics index you use for your project must be active and have queries enabled for your project to function properly. Before completing a full or incremental population of your index during an ongoing project, we recommend turning off all review queues and turning them back on once the index is active. Project reporting may be incorrect during a full or incremental population but will be corrected once the index is active.

Reviewer group

Create a reviewer group with the users you want to access the Active Learning project. You can add or remove users from the group at any time.

Note: The users in this group are not automatically added to the Active Learning project. You must grant each individual access to the Review queue. For more information, see Project setup.

This review group must have the following security permissions:

Object Security Tab Visibility Other Settings
  • Document - View, Edit
  • Documents
  • Browsers - None OR Folders and/or Field Tree and/or Clusters
  • If Browser permissions are set only to Advanced & Saved Searches, reviewers can't access the Reviewer page.

For more information, see Workspace security.

Review field with two choices

Create a single choice field with two choices for reviewers to code on. One choice must represent the positive/responsive designation, and the other the negative/not responsive designation. For more information, see Fields.

Note: We recommend turning off family propagation with Active Learning.

Review layout

Relativity will not create a layout automatically; you must create a layout for reviewers to make coding decisions on. However, this layout is not a prerequisite for creating your Active Learning Project. For more information, see Layouts.

Pre-project sampling

The Active Learning model only builds once you have at least five documents coded with the positive choice and five coded with the negative choice. If there are fewer than five coded documents for each choice, the system serves random documents in the prioritized review queue until the threshold number of documents are coded. You can use pre-coded documents to help start the model's learning and begin the project with a model build and a ranking of all documents.

If pre-coded documents exist, you can use these coding decisions to start the model’s learning. The prioritized review queue can then serve up the highest ranked documents for review. You must ensure these coding decisions are set on the project review field, and that at least five coded documents exist for both the positive and negative choice.

If no pre-coded documents exist, you may want to have reviewers code a sample of documents prior to starting the Active Learning project. You can do this either before or after you create the project. You can draw a sample of documents with specific keywords, from key custodians, within a certain date range, etc. to help focus the sample on documents more likely to be important in the case. You must ensure these coding decisions are set on the project review field, and that at least five coded documents exist for both the positive and negative choice.

Note: If relevant documents are very rare, then targeted searches for relevant documents outside of the queue can help make Active Learning more effective.