FREE NEWSLETTER

GET EXCLUSIVE CONTENT

Free Newsletter Free Newsletter
reveal-banner-180px-x-150px-v1gif

ThoughtRiver’s founder tells us why they’ve won a £400k award for the development of ‘thinking AI’

Added on the 13th Dec 2018 at 11:45 am
Share Button

Contract intelligence vendor ThoughtRiver has been awarded funding from Innovate UK for a £400,000 project to develop what it refers to as ‘thinking AI’.

ThoughtRiver’s technology automates the day-to-day routing and prioritisation of decisions in corporate legal functions, using machine learning to interpret contracts and assess their risk based on corporate policies and protocols.

Each contract pre-screen undertaken by ThoughtRiver triggers 1000s of software-answered questions on each contract in order to replicate what actually happens when a junior lawyer or paralegal conducts an initial scan on a contract.  Pre-screening then rolls up the answers in order to deliver a single final context-sensitive recommendation about whether a contract needs further review and by who.

The new research project, which ThoughtRiver will undertake in collaboration with NLP consultancy iLexIR and its founder Professor Ted Briscoe of Cambridge University, is intended to result in a significant advance in the ability of computers to rationalise complex problems in novel scenarios based on self-taught logic.

Speaking to Legal IT Insider, ThoughtRiver founder Tim Pullan said: “Technology like ours is looking to save the user time in finding out information locked in a contract so that they can quickly move to the next task.

“Obviously a lot of the technology today is effectively searching for things that have a certain meaning and our system works differently because we abstract it and roll it into a single decision but nonetheless, we use similar technology to summarise the contact.

“However smart that is, what clients really want to do is in a few simple words to ask a very complex question. You might ask whether the contract limitation of liability is limited by reference to the contract value, and that’s a question you can ask with today’s technology. But if you were to ask ‘is the liability in this contract more than £1m?’ and the contract says ‘the liability is not greater than 100% of the value of the contract’ then to work out if that is more than £1m, a human would have to hunt to find the value of the contract and work it out.

“If this was your own standard contract you could use rules to automatically extract that information but in the real world it’s not about the contracts you know but the ones you don’t know that are the issue and if your system is confronted with a contract that it hasn’t had specific rules designed for, then it can’t give you the answer. What we are doing is developing technology that will have the possibility or the capability of giving you that answer.”

Professor Briscoe is well known in the field of computational linguistics and an inaugural director of the Alta Institute, which conducts research into automated language teaching and assessment.

Innovate UK, meanwhile, is a competition designed to encourage companies like ThoughtRiver to invest in cutting edge development and comes as the UK government takes an interest in investing in legal technology to retain a competitive edge on the world stage.

Pullan said: “Entries are judged on technical merit and commercial application – they are not going to fund research that is naval gazing; it’s to do with having an idea that is a great idea and can commercialised.”

In a statement co-chair of APPG AI Lord Tim Clement-Jones said of the Innovate UK award: “I am a strong advocate of world-leading AI R&D taking place in the UK – especially in the context of key export industries such as legal services – and thrilled that Tim and the ThoughtRiver team have been awarded this project”.

Professor Briscoe added: “This is a technically exciting and challenging project and we are looking forward to working with the ThoughtRiver team in 2019.”

2 Comments

  1. Alex Smith says:

    This sounds very familiar to reported projects, research and features around Lexis Answers and West Answers … looking to find contextual ‘answers’to questions posed of a system. Lexis Answers info here https://www.lexisnexis.com/infopro/keeping-current/b/weblog/archive/2017/06/29/you-ask-lexis-174-answers-new-machine-learning-feature-on-lexis-advance.aspx and various interviews with Jeff Pfeifer in the media this year.

    It also reminds me of the answers work of Dan Arrowsmith at Tolley (play with it here https://www.tolley.ai) and article on it here https://www.artificiallawyer.com/2018/09/04/meet-tolley-ai-the-worlds-first-tax-bot-and-help-it-to-learn/

  2. Tim Pullan says:

    hi Alex, good to hear from you. I am not familiar with Lexis Answers or the other systems here but reading the links you provide what we are doing is very different. This is not contextual search or focused on the tech around parsing questions. Unfortunately the technical details are confidential but happy to brief you when the project has completed! Cheers Tim

Any Comment?