The potential to improve internal processes

But there is other valuable data within those contracts that, while it may not impact revenue, provides important information for other functions. Addresses and contact information help the billing team get invoices out to the right people, on time. Visibility into limitation of liability caps or insurance requirements provides actionable insight for legal. The ability to extract data beyond the bare minimum required for revenue recognition, without additional manual work, has the potential to exponentially improve internal processes and cross-functional collaboration.With Klarity, the number of documents you can review goes up, and the data you can extract expands in kind. We integrate with core systems like Salesforce and Netsuite to pull all relevant documents into one central repository. From there the document goes through a series of steps during which it is broken down in various different forms to automatically extract key data.

‍

‍

First, we use optical character recognition (OCR) to extract formatting and structural information, like bold, italic or underlined text, which provides context around how the contract is structured. Semantic and linguistic features are also identified and automatically extracted. Once a document has gone through pre-processing and its key data points have been extracted, our model – which was trained to identify important rev rec concepts like a termination for convenience clause – allows the revenue recognition team to  to automate their review.

‍

‍

The next step segments each document into various parts. Segmentation breaks down the document into semantic units that will later be classified as part of a particular concept, or not. We attribute features to each of these semantic units, analyze the syntactical layer, and perform a semantic evaluation of each word within its original context. This allows Klarity to identify concepts of importance, language related to termination, indemnification or liability. The system is pre-trained to identify more than 340 key concepts and categories that it automatically searches for before a human ever has to review the document.

Klarity uses pre-trained language models along with those 340+ concepts to review every word in the document and extract key rev rec information. Klarity then extracts metadata related to key contract information, like parties involved or effective date, providing another set of models that run with the same input as the features used in the previous step. These models automatically extract all relevant information into the parties involved — primary vs. secondary, key dates, and more.

Finally, the system performs what we call “clause extraction” — information like whether a termination clause is for convenience or cause. Klarity looks at all of the features it has seen so far, which are present in each document and where they are located to determine which core concepts need to be flagged for human review. Klarity extracts hundreds of different clauses across a variety of document and contract types, including MSAs, order forms and purchase orders. Each different contract type has its own specific core concepts most critical for extraction, and Klarity uses a union of these different models to help identify the right ones.

As your organization grows and regulatory thresholds are raised, it’s important to be more comprehensive about what you look for. Simultaneously, the need for efficiency becomes paramount. Klarity gives you the power to continually extract valuable data as the volume of documents that require review rises.