Heretik Analysis Part 2: Extraction & Unitization

Heretik Analysis Part 2:
Extraction & Unitization

Rishi Khullar | June 6, 2019

If you’ve been following us recently, you’ll know that we have been working on something pretty huge. Heretik has recently announced our newest addition to its growing toolkit: Heretik Forge.

Forge was built on the basics of Heretik Analysis, which include Segmentation and Classification. We talked about these in our Heretik Analysis Part 1 blog post. If you’re curious about these types of analysis, make sure to check out that post!

But, if you’re a Heretik regular or you’ve already read that post, buckle up. We are about to dive into what we’re working on now, and how it will help our customers get to structured contract data faster and more efficiently.

So first, a quick refresher:

Segmentation Analysis allows you to dissect a document or contract by breaking it up into specific sections.

Classification Analysis can arrange contracts and sections into categories. (ex. Licensing Agreement)

These are the foundation of Heretik’s Analysis, but there are two more that will be added. Extraction and Unitization which work in tandam.

Building upon the basics, Extraction Analysis does what it sounds like. When running Analysis, it can be used to extract important and particular information from a contract. For example, the focus of your extraction may be the fallback language for a contract that has LIBOR. Extraction Analysis will pull out this information and populate it to a Relativity field.

We are attacking Extraction Analysis from two angles:

Providing you a tool kit to craft custom regular expressions to find specific information and auto-populate Relativity fields
Out-of-the-box machine learning models that identify specific information and auto-populate them to Relativity fields that Heretik will ship with (e.g. Heretik Effective Date).

For #1, we built this as part of the May Release and customers are already saving weeks of review time taking advantage of this functionality! For #2, our engineering team is underway implementing the hard work our data science team has done to build models for extracting effective dates and parties, to start.

Unitization Analysis then comes into play by piggy-backing off of Extraction Analysis in order to auto-group related documents such as amendments with master agreements. More often than not, there are certain hints or signals that indicate such a relationship between documents. Unitization Analysis will help attack the painful problem of having to manually group related documents before you can kick off a review project.

Chis Tkach, Senior Client Success Manager at Heretik, has over a decade of experience managing contract review projects at Axiom. “A lot of times clients will say they have 10,000 documents — a mix of master agreements and amendments, SOWs, purchase orders, etc. (ancillary documents), but they don’t know the exact number of contractual families,” says Tkach. “In order to properly scope a project for budget and timeline, it’s imperative to know the makeup of your corpus of documents. You could simply assume a number of masters agreements and ancillary documents and then do a metadata review, which could take weeks and costs tens of thousands of dollars, in order to confirm the exact number. If you think it’s 2,500 masters and budget and staff for that, when you find out it’s 5,000 masters, you’ll have to go back to client with increased budget and extended timeline. With Unitization Analysis, you can know the number of contractual families in hours and therefore efficiently price and staff your projects.”

Unitization will be integrated into Heretik Forge in the fall of 2019 and our first out-of-the-box models for Extraction Analysis will be available in the summer of 2019. To find out more information, don’t hesitate to email us at product@heretik.com!