What You Need to Know About Continuous Active Learning (CAL) & Automation Part II
On May 30th, we held our third webinar in our ongoing series this year of webinars about new technology innovation in the eDiscovery space. This webinar was focused on the latest in Continuous Active Learning (CAL), a hotbed topic for attorneys and organizations today. Our webinar focused on sharing our definition of Continuous Active Learning (CAL) and how it fits into how clients using VenioOne Review will further benefit from our latest enhancement to our discovery workflow and review automation.
A High-level Overview of the Continuous Active Learning (CAL) Workflow:
- Create CAL Profile and Review a document set.
- Reviewers then review batches.
- CAL Engine is working behind the scenes automatically.
In this webinar, Venio’s CEO/CTO, Arestotle Thapa, approached the new technology from two viewpoints, a demonstration from a reviewer’s perspective, and the review administrator’s perspective. To wrap up, he demonstrated how the VenioOne CAL engine works in the background to complete the review.
In the webinar, we shared: The differences and benefits of Continuous Active Learning (CAL). Unlike VAR, training documents in CAL are coded by Reviewers, not SMEs. The CAL training module engine is continuously being updated based on Reviewer’s coding decisions, not stagnant. CAL Review and Training are the same processes. Meaning: It lowers costs, is less effort, and it gives you faster results and relevant documents.
From a Reviewer’s Perspective: Your reviewers are checking out the documents assigned to them, they make decisions on if the doc is relevant to the case. Once they finish the batch of documents and check them back in, the reviewer receives another batch and starts the process over until there are no more documents.
How CAL Works as a Reviewer:
Look at the document, flag any type of document as relevant, and the reviewer looks at the document to decide if it is relevant, and moves on. Can be done on a doc by doc basis, or in bulk. A user can also set specific flags. Users can individually mark documents as not relevant and group documents in batches.
In the beginning, from CAL, you’ll have a smaller number of relevant docs, and as CAL learns, you’ll get richer batches of relevant docs, and then eventually all the irrelevant docs will be filtered out.
CAL from the Perspective of a Reviewer Admin:
An Admin creates a CAL profile. Your CAL profile gives instructions to the CAL engine on what it needs to do during the review process behind the scenes. As part of creating the profile, you tell CAL what tags it needs to apply and what categories it should learn. You can also tell CAL what percentage of documents you need to be relevant. This means that CAL will continue sending documents to the Reviewer until it meets that requirement or threshold. We recommend each user tell CAL what percentage of each document needs to be relevant, and even if you hit your goal, CAL will continue sending documents until the percentage of relevancy drops below your set amount per documents.
For example, in CAL a user can say “here are the categories I care about, and these are the thresholds to looks out for.” A user can also define how the reviewer manages to redact certain things or not.
Then, you log in and start to review the documents. Your dashboards can give you different reports on how CAL is learning and how it predicts the number of relevant documents, how far along in the documents they are, and how the reviewers are performing.
CAL is Designed to be User-Friendly & Efficient
To make the platform as user-friendly and efficient as possible, Venio offers you templates for profile creation. If you want, you can tell CAL what type of document to lean more towards, which will present itself as a tag just like any others. In addition, you can set which users are your Reviewers, and assign what data they can see. Your admin dashboards can show you how the predictions are working, how the review is going, and what the progress is of the review. We designed CAL to provide easy to understand charts that visualize progress and performance.