Clausematch has released a digital knowledge graph in open source which aims to drive AI-powered digitisation across the regulatory industry.
The London company, which topped our RegTech 50 ranking last year, says the graph taps into a dataset created with regulators and uses visualisation to fundamentally transform the understanding of regulation and how it can be delivered.
Regulators and financial services companies now have access to test the graph and see how regulation in a structured digital format works.
Clausematch has been involved in a series of projects in this domain with financial services regulators since 2019. The aim is to take the content of regulatory requirements, automatically categorise them using advanced AI models then create tags focused on regulatory concepts, obligations and expectation.
Evgeny Likhoded, CEO and founder, said that with a regulator such as the Financial Conduct Authority there might be hundreds of pages of regulations for a startup to sift through as they attempt to find the passages relevant to them.
By training the model to think like a regulator, these companies can then “see their own version of the handbook for their industry… or map their own”.
He added that, given the general apprehension around machine learning and AI, actions taken by the algorithms are fully open and transparent.
“We need to get to a point where regulators are comfortable with the machines and their accuracy,” he said.
Eventually, it is hoped that the graph will enable more institutions and firms to foster greater governance practice and transparency in their daily operations, being compliant by design.
“There is a growing sense that authorities need to work together to regulate the world more effectively, learn from and leverage each other’s experiences, and avoid silos of expertise and skills,” said Vladimir Ershov, head of data science and machine learning.
“By introducing dynamic knowledge graphs in open source, Clausematch proposes an approach where the regulator first digitises its very understanding of regulation by creating models for tagging and for relation extraction and then, these models are released within the regulations for regulated companies.
“Thus, any company can receive an in-house regulatory advisor to carry on with taxonomy and rules interpretations.”
In Crypto: UK government to publish draft regulation rules today
Likhoded added that “regulators need to become active advocates of a new structured approach to regulatory texts” and that such an approach can also enable parties to react with speed to changes in regulation.
“When regulations are digital, it gives a tremendous advantage and capability to businesses to achieve an entirely new level of compliance and transparency,” he said.
“Working with regulators and financial services institutions, we are committed to continue to innovate and take compliance to the next level. The more regulators align with this vision of structured machine readable regulations, the faster we can propel the industry to improve and significantly reduce the cost of compliance.”
Clausematch has published a scientific paper on the application of knowledge graph technologies for regulations.