Zuva, the document intelligence company that span out of Kira Systems, has launched an AI Trainer capability, which allows the rapid building of NLP models without needing help from technical people. Below is an AL TV Product Walk Through of how it works.
The AI Trainer allows a user to train the system to find specific text or particular clauses, and soon whole document types. This lets any user, without having any experience of NLP tools, to quickly build models to find the information they want from a doc stack.
In the demo, CEO, Noah Waisberg, and Resa Jacob, Legal Knowledge Engineer, at Zuva, explained to Artificial Lawyer that once you get 10 to 20 examples of what you are looking for into the system – which you do by highlighting the item you want to find elsewhere – then Zuva starts to auto-suggest other examples, which you can then approve or reject, thereby improving the model.
Waisberg added that Kira had the first no-code machine learning contract training interface for NLP, and naturally Zuva’s new AI Trainer builds on that experience.
He noted that what it offers is significant for three reasons: ‘It shows that we’re making fast product progress here at Zuva. We only spun out back in September 2021, and it’s not easy to build something like this quickly.
‘AI Trainer means that subject matter experts can train models without needing a developer or data scientist to intermediate. [And] this is pretty unique among embeddable contracts AI tools. With comparable other tools, our impression is that a technical person inputs examples into a system, and the system spits out something for a technical user. With AI Trainer, a lawyer can just highlight text and the system gets to work behind the scenes.’
So, here it is. The AL TV Product Walk Through is about 10 mins. Press play to watch inside the page.
Waisberg also noted that they had taken the decision to publish their pricing and to allow potential clients to have a free, limited, use of the system to try it out. He added that although customers are welcome to contact them for input, they encourage new users to just come to the site, get set up online, and get right into doc analysis projects.
The key point here is that some of the most scalable software businesses are those that enable frictionless customer experiences, without the need for a lot of additional input from the vendor.
Is this a big deal? Overall it should make Zuva relatively easy to use and speed the process of getting results, which will be important here as Zuva is designed to be used across multiple areas in a company, not just legal, and so needs to be very adaptable as customers work on a wide range of use cases and doc types.