Listed data and software company Nuix has been granted a patent for pioneering a novel technology system that uses advanced deep learning techniques to better train artificial intelligence models.
The newly patented technology tackles one of the most complex challenges of neural language processing in AI systems – accurately classifying documents that share significant similarities in language and structure but belong to distinct categories.
For example, distinguishing between non-disclosure agreements and non-compete agreements, which often overlap in terminology, has historically been difficult for automated systems. Traditional approaches struggle to achieve accurate results, require time-intensive retraining when new categories are added, and often result in false positives or reduced classification accuracy.
The invention also eliminates the need to train separate neural networks for each category of documents, making it highly efficient and scalable for organisations looking to automate their workflows, reduce operational costs, and derive actionable insights faster.
“Earning this patent is a fundamental breakthrough and a credit to our hardworking and dedicated team,” said Nuix CEO Jonathan Rubenstein. “We’ve created a system that revolutionises document classification in natural language processing, improving accuracy and reducing false positives.
“The patent underscores our commitment to innovation, while expanding our portfolio and potential revenue streams,” he added. “Innovations like these are proof that we are not only keeping pace with AI developments but embedding ourselves as the picks and shovels of the AI revolution.”
Nuix’s technology holds wide-ranging applications across industries that rely on accurate document classification, including legal, compliance, forensic investigations, cybersecurity and data governance sectors. Enabling organisations to efficiently and accurately process and categorise documents allows for better decision-making and operational efficiency.
“By addressing the limitations of conventional neural language processing methods, we’ve developed a system that not only improves accuracy in document classification, but also substantially reduces training data requirements and streamlines workflows for our clients,” said Nuix Chief Field Technology Officer Stephen Stewart. “This breakthrough in advanced data analysis strengthens our role as a trusted technology partner for industries that demand precision and efficiency, from legal services and compliance to cybersecurity and forensic investigations.”
