Learn more about
Heretik Analysis
Heretik’s number one objective is to provide our clients a toolkit that allows for the right amount of flexibility within a repeatable and scalable framework to tackle large document review projects.
This approach has been used by the eDiscovery industry for over a decade and is one, thanks to Heretik, that you can easily apply to regulatory response, corporate transaction, and contract data management use cases.
From day one, we have focused and prioritized these core functions with-in our software:
Document Review
There is no easy button for getting to truly structured contract data. It’s a combination of machine learning and a best-in-class document review experience. This last pillar has been missing in the market for years in contract review but has been evolving rapidly in eDiscovery with Relativity at the forefront. Heretik is the only contract review solution that is built on top of a leading document review platform that excels in security, scalability, and configurability.
Truly Structured Data
Knowing which contracts that are coming up for renewal in the next six months should be as easy as clicking on a pie chart or filtering a list. Any analytics tool that solely identifies blocks of text in a contract does not get you here. Heretik is the most complete and end-to-end solution for efficiently transforming contracts into truly structured data in the form of robust, comprehensive, and discrete data types.
Collection and Processing
In a world of rapid digitization, tomorrow’s worries are quickly becoming today’s crises, resulting in companies continuing to struggle with digital organization. Easily manage the collection, identification, and processing of vital documents scattered across multiple repositories with the help of forensic teams.
Speed to Review
The sooner teams can begin analyzing documents, the sooner expert problem solvers can add their contributions. Teams no longer need to wait for training rounds or seed sets to begin analysis. Start working with the documents in minutes and respond to the results in real time without needing new data sets or models.
Heretik Analysis is comprised of the following functionality:
- Contract Classification
- Section Segmentation and Classification
- Data Extraction to Fields
- Imaging and OCR
Contract Classification
This pre-trained model identifies and auto-populates the contract type for each document to a single choice Contract Type field. Heretik will also populate the verbatim contract title to a fixed-length text Contract Title field. In tagging each contract with a type and title, Heretik enables you to understand how many leases, employment agreements, or vendor agreements you have, for example, and gives you the ability to filter, folder, prioritize, and otherwise organize a corpus of contracts.
Section Segmentation and Classification
This pre-trained model identifies over 330 types of sections (or clauses) within a contract and performs three critical actions:
1. For each section found, Heretik creates a new Relativity document and auto-groups it to the parent contract. With all sections as their own documents in Relativity, this empowers you to compare the wording of a confidentiality section, for example, across all contracts and take action by coding at the section level. This way, “show me all confidentiality sections that need re-papering” becomes a simple click to filter.
2. For each section found, Heretik will auto-populate the following fields:
- Section Heading – a fixed length text field containing the verbatim section heading from the contract.
- Section Type – a single choice field that denotes the type or category of section, such as Confidentiality, Governing Law, or Indemnification, as examples.
3. For each parent contract with sections, Heretik ships with a script for you to populate and keep evergreen the following fields:
- Number of Sections – a whole number field providing the number of sections found in the contract.
- Sections Types Found – a long text field providing a roll up of all section types found in the contract.
- Section Headings Found – a long text field providing a roll up of all section headings found in the contract.
If you have contracts that are highly bespoke such that a segmentation model can’t identify sections, you can use Regular Expression Segmentation to write your own search pattern to segment the document and bypass the need to spend time and money training a model, allowing you to begin the review for deeper insights sooner.
Data Extraction to Fields
While identifying contract types, section types, and blocks of text that comprise sections is highly valuable, in many cases it’s a means to an end of diving into deeper analysis and unearthing more granular risks, obligations, and opportunities in contracts. The final output of this deeper insight MUST be structured data in the form of fields and choices so that clients can visualize and report on this critical data (reporting that’s not possible with mere text blocks).
You can conduct this deeper analysis manually in the Heretik Viewer by using our robust search & navigation capabilities along with Send to Field, but you can also automate a large portion of this field population via our Data Extraction to Fields analysis.
With this type of analysis, you can create, customize, and tweak an advanced search pattern and then map results of that search to your own custom fields. For example:
- There are 10 or so different variations of “construed by the laws of”.
- You could easily tell your search to look for a state that is mentioned shortly after one of these variations.
- You could also tell Heretik to only populate the state name to a field called Governing Law and nothing else that the search returned.
- Also, you can tell Heretik to ignore all results of the search if it does not hit in a section of type Governing Law
- The end result could be the Governing Law field populated automatically, and in a highly accurate way, across thousands of contracts in your data set for a QC team to review.
Imaging and OCR
We’ve developed our own OCR that runs during Heretik Analysis so you no longer have to run this process beforehand to populate extracted text. We’ve built on top of an open source OCR engine and utilized numerous techniques to modify and tailor our OCR to the use case of contract review. Now that we control more of the end-to-end process around structuring contract data, we position ourselves to cater our OCR specifically for documents like contracts and attack painful OCR problems such as double columns, tables, and readability. Hererik OCR is designed to complement your existing toolkit, not replace it. Here is how it works:
- If the Extracted Text field is populated, Heretik will NOT run Imaging and OCR and will instead analyze the document based on the text in this field. This way, if you’ve already run OCR on documents and are happy with the results, you can avoid added time that Heretik OCR would take to image and OCR the documents.
- However, if you have run OCR on documents and populated the Extracted Text field and were not happy with the results, you can try out Heretik OCR by simply clearing out the Extracted Text field and running analysis on those documents.
- When running Heretik OCR, we will auto-populate an OCR Confidence Score field to give full transparency into the quality of the OCR so you can better prioritize review.
