What Is Machine Learning (ML) and Natural Language Processing (NLP)?

To make contract discovery and analysis work accurately and efficiently, there are several technologies that need to be used together. This includes NLP, which uses complex rules sets and statistical analysis for the identification of terms and pro­visions in the discovery and initial data extraction, and ML, which uses combi­nations of algorithms to support the creation of custom extraction policies and trains the system on the specific language, products, provisions, and clauses for a parti­cular organization.

The system learns through examples, processing multiple contracts over time and receiving feedback on how it identifies various components of a contract. As a consequence, the system becomes more effective and increasingly precise in finding what it is seeking. These systems don’t just “find,” using keyword search or text matching, but they “think,” using combinations of words and phrases to deduce concepts and gist, and then they are taught if they are right or wrong, and that information goes into the learning aspect. This is where the idea of intelligent machines emerges.

There are also new techniques in the field of AI called “deep learning” that uses Neural Networks in much the same way as the human brain. One form of Neural Network is called Convolutional Neural Networks (CNN). By using CNN it’s possible to reduce the computing power required (i.e. reducing the size of the “brain”) and increase the speed at which the system can learn, as a key element of any system is its ability to learn quickly and accurately. This is critical when large organizations sorting through 10,000s to 100,000s of documents to find important pieces of information.

As discussed, it takes the right combinations of NLP, ML, the right set of algorithms with technology that makes them all work together. Important to all of this is the data visualization platform which will present information intuitively in order for a system to work properly.

It is interesting to see how Intelligent Contracting plays out in practice. Take for example “Change of Control” provisions. Change of Control is a critical field in all mergers and acquisitions transactions, but the provisions can easily hide in assign­ment or termination provisions, be their own distinct provisions, or hide in plain sight with non-standard language. Take the following examples of Change of Control provisions:

Assignment. Company shall not assign, delegate or otherwise transfer the agree­ment or any of its rights or obligations under the agreement without prior written consent. A Change of Control shall be considered an assignment for the purposes of this provision. Any non-conformance with this section shall constitute a breach and immediate termination.

Change of Control. In the event of a reorganization, merger or consolidation that results in the transfer of 50 %o or more of the voting authority or a sale of all or substantially all of the assets of the Company to another person or entity, this contract shall terminate.

Terminationfor Reduction in Ownership. In the event that Company's Aggregate Ownership Interest is less than five percent (5 %), Client shall have the right to terminate this Agreement immediately by giving ninety (90) days' written notice of termination to Company.

To add another wrinkle, let's consider Change Control language. Change of Control provisions are much different than Change Control language which often describes the process of how to change the type, nature, delivery, etc. of work define/described in a Statement of Work (SOW) or similar type contract. This process can be pages and pages long, so the below is significantly consolidated for illustrative purposes:

Change Control. Should there be any changes suggested under this statement of work (SOW), those changes shall be considered out of scope and an amendment to this SOW or a new SOW shall be required.

Now the question becomes, how does NLP and ML help with these variations? Imagine you are the General Counsel of a highly acquisitive company, and after multiple acquisitions, you recognize that many of your contracts have Change of Control provisions and you need to determine which contracts are potentially terminated due to the new company structure. You have prioritized your supplier agreements as the starting point to begin to determine where Change of Control comes into play. The challenge is, you have over 10,000 suppliers with each supplier having over 20 contracts associated with the relationship, so essentially 200,000 contracts to sort through. The options are to hire a team of lawyers to begin the sorting process (and finding the budget to do so), teaching the software to find the contracts you require, or do nothing and hope for the best.

Since “doing nothing” is not an option as legal counsel during an acquisition, only the first two options are viable. Here is how the math will break down for hiring a team of lawyers to look through your contracts:

Outside Counsel Fees

Legal Process Outsourcing Resources

Now, consider teaching software to do the manual process of sorting through the contracts so the lawyers only need to review the contracts that contain the language, and the language includes the various standard and nonstandard language, yet excludes the Change Control language because the software recognizes that the intent of the Change of Control language is different, even if some of the proximity of the words is similar.

What you’re left with is under 10,000 contracts that can be sorted by those that may terminate in order to begin the renegotiation process or tagged as “inactive”. This type of sorting can typically be managed in-house and only requires the cost of the software.

In-house legal departments whom can articulate this of result is much more likely to be allocated budget for new technology investments because the return on investment is clear.

5