Diligen launches ML-based self-training system

Diligen today (29 July) announces the availability of Diligen Prodigy, a new self-training system that will enable law firms and legal departments to more rapidly train Diligen to recognise new clause types in contracts.
The system learns to recognise new concepts in minutes. If, for example, you want to teach the system to look for recital ‘now, therefore’, you can start with a keyword search and a simple ‘yes’ or ‘no’ to the data points presented. In the demo we were given, it was fairly effortless.
Rather than having to painstakingly dig through documents and highlight the relevant text, Diligen Prodigy volunteers the data for the user to classify. Affirmed clauses appear under ‘yes’ on the left of the screen and discarded ‘no’ clauses on the right, complete with who worked on them. The clauses can be dragged and dropped from one pile to the other.
COO and co-founder Laura van Wyngaarden said in the demo: “Every single data point will update the ML model. We have dramatically simplified this process so that you’re inputting 10 initial examples through a search or pasting. Then our machine learning is doing a lot of the heavy lifting to generate suggestions: it looks at what you’re providing and changing in real-time to provide different types of data points very simply on one page. The user is making a very easy determination about whether to include that data piece in what you’re training and you’re getting real-time feedback on the performance of that model.
“The old way would require digging through documents and highlighting paragraphs and a lot of iteration, instead of saying ‘yes’ or ‘no’ – this is a huge departure from how these systems have traditionally worked.”
You can pretty quickly start to see the system serve up clauses it knows are recitals.
We’re not suggesting that this is the only system capable of self-training, but Friedrich Blase, an advisor to law firms/law departments, conducted a side by side review of 12 AI-based contract review tools, and estimates that roughly 20% of the available solutions currently offer self-training.
“When I conducted the comparative trial of 12 contract extraction tools for a consortium of legal departments, they struggled most making do with pre-trained clauses and applying them to their specific use cases,” Blase said. “The new self-training capability in Diligen’s solution not only shows how fast technology is evolving but also gives law firms and legal departments new immediate/do-it-yourself options which increasingly are in high demand.” We’ll try to get you more detail on that study.
Founded in 2015 in Toronto in Canada, Diligen’s core business is machine learning and contract analysis, focusing on everything from due diligence to lease extraction, compliance, contract lifecycle management, data migration, metadata extraction, audit, finance, and corporate documents.
It works with corporate legal departments, law firms, ALSPs and a yet to be announced Big Four accounting client, so watch that space.
“As a South African Legal Services Provider, Inlexso (Innovative Legal Solutions) has been faced with “teaching” machine learning based contract review systems with bespoke clauses unique to the South African legal landscape,” noted Cliff MacGregor, Managing Director, Inlexso. “Having used multiple machine learning contract analysis systems, we found that the Diligen training model and process are user-friendly and efficient. We believe that this enhanced efficiency will improve project execution time and accuracy, ultimately resulting in time and cost savings.”
For more information about Diligen Prodigy or to set up time to see it in action, go to https://www.diligen.com/demo/