Project pipeline steps

Each aiR for Privilege project has a dedicated set of pipeline steps that must be completed in sequential order to generate document results.

Note: Pipeline steps should be performed by a Case SME. For more information see Roles and security permissions.

The pipeline steps are as follows:

  1. Prepare Project
  2. Scrub Disclaimers
  3. Identify Spammers
  4. Classify Domains
  5. Match Equivalent Domains
  6. Validate Attorneys
  7. Confirm Privilege Status
  8. Populate Privilege Results

The Action buttons on each pipeline step let you both Run and Apply Annotations for each pipeline step. Most pipeline steps function similarly:

  1. Perform a Run action to kick off the AI to analyze and identify certain content.
  2. You and your team annotate the predictions made by the AI. See Perform annotations.
  3. Perform an Apply Annotations action so the AI can learn where it was correct and wrong to make final calls and act on them.

For example,

  1. You would Run Scrub Disclaimers to locate all potential disclaimers.
  2. Confirm that each Disclaimer identified is actually a Disclaimer by performing annotations.
  3. Use Apply Annotations to remove disclaimers from the analyzed text to reduce false-positive privilege hits.
Note: The Prepare Project and Populate Privilege Results pipeline steps do not require any annotation, while the Confirm Privilege Status pipeline step does not require any AI analysis.

Pipeline step throughput benchmarks

Following are estimated times for the system to complete a Run action and an Apply Annotations action for each pipeline step. We gathered these numbers from an average 50,000 document project.

Note: These numbers represent one example and can vary based on the unique attributes of the dataset and aspects of the system. Note that if you are trying to estimate run times for larger projects, the times listed do not scale lineraly, but rather logarithmically. More exact throughput metrics will be added to documentation before general availability in Q3.
Pipeline step System Run Time System Apply Annotations Time Notes
Prepare Project 58 minutes    
Scrub Disclaimers 3 minutes 1 minute  
Identify Spammers 1 minute 1 minute  
Classify Domains 5 minutes 1 minute  
Match Equivalent Domains 25 minutes 1 minute Run time for this step can vary greatly.
Validate Attorneys 43 minutes

Round 1: 200 minutes

Round 2: 15 minutes

Round 3: 15 minutes

Round 4: 15 minutes

 You apply annotations four times for this step.
Confirm Privilege Status   1 minute  
Populate Privilege Results 15 minutes   Once this step has completed, Privilege results will begin to trickle in across documents. All results will populate in about 35 hours.

Pipeline step notifications

You can specify email addresses that will receive notifications when a pipeline step's status changes by entering Notification Email Addresses. You can enter Notification Email Addresses when creating a project or add them when a project is in progress.

To add Notification Email Addresses to an existing project:

  1. From the Privilege Project page, select Settings.
  2. Select Edit.
  3. Add email addresses that should receive notifications to the Notification Email Addresses (comma delimited) field.
  4. Click Save.
    An image of the Settings page and the Notification Email Addresses field

Abandon a project

Only one privilege project can be In-Progress at a time within a workspace. If a project becomes blocked during any of the pipeline steps it may be necessary to abandon the blocked project and start a new one.

To abandon a project:

  1. Navigate to the Projects (Priv) tab and select a project from the list.
  2. Click the Abandon button at the top of the Privilege Project page.
  3. The Abandon Project pop-up will appear. Click Abandon to confirm that you want to abandon the project.
    An image of the Abandon Project pop-up
Caution: Abandoning a project cannot be undone.
Note: If the Prepare Project pipeline step was successfully completed before abandoning the project, you were already billed for the document units included in the project.

Once you abandon a project its status will change to Abandoned and you can no longer perform pipeline steps for that project. The system unlinks any documents associated with the project so that you can run them through a new project.