Developed at the University of Cambridge, Luminance’s core intelligence is the Legal Inference Transformation Engine (LITE). Exposed to over 100 million documents so far, Luminance’s language model is able to rapidly read and form an understanding of documents, before clearly displaying the results of this analysis to the user. Luminance’s AI combines a number of disciplines within the field of machine learning, including inference, deep learning, natural language processing and pattern recognition. Importantly, Luminance is unique in combining both supervised and unsupervised machine learning, giving lawyers the most robust, powerful platform for legal document analysis and process automation.
Developed at the University of Cambridge, Luminance’s core intelligence is the Legal Inference Transformation Engine (LITE). LITE is able to rapidly read and form an understanding of documents, before clearly displaying the results of this analysis to the user. LITE combines a number of disciplines within the field of machine learning, including inference, deep learning, natural language processing and pattern recognition. Luminance is unique in combining both supervised and unsupervised machine learning, providing lawyers with the most robust, powerful platform for document analysis.
Unsupervised machine learning does not require any a priori knowledge of labels or tags in order to solve a problem. Instead, the machine is exposed to vast datasets and by analysing these in their entirety, the underlying patterns of the data are revealed. The system quickly builds up a pattern of what is normal – and thus what is abnormal.
Only unsupervised machine learning, which does not rely on understanding documents in relation to given labels, surfaces the ‘unknown unknowns’ – issues that the reviewers did not know existed and thus never searched for them or labelled them, but that nonetheless present as a significant anomaly. This is critical in due diligence as all risk is surfaced immediately within the review and negotiating positions can be strengthened from the outset. Similarly in eDiscovery, unsupervised machine learning can help litigators quickly find the smoking gun amongst mountains of irrelevant material.
Within supervised machine learning, the system is exposed to examples of data that are described and defined. The system forms an understanding of what makes up each classification, and can apply these to new datasets that it sees.
In fact, Luminance’s technology takes supervised machine learning a step further, as it continues to learn from the ways that lawyers interact with the platform, too. By observing the actions of legal professionals, Luminance can form an understanding of what lawyers are looking for in their documents, meaning that Luminance actually becomes more intelligent and bespoke for each lawyer the more that they use the platform.
AI-generated language models are becoming increasingly sophisticated in their ability to understand the written word. Establishing an understanding of a diverse range of parameters is of critical importance when building language models, since the scale and variation of datasets that the technology is exposed to will ultimately define the quality of output. Indeed, building neural networks at scale requires vast data flows: Luminance has now seen and analysed over 100 million documents in over 80 languages across hugely diverse fields of law, from Non-Disclosure Agreements and property leases, to emails and Whatsapp messages. This makes Luminance the most advanced language model accessible to lawyers today, ensuring lawyers can be confident that potential risks, anomalies or contentious issues are surfaced when using Luminance’s AI to draft and review documents.
Please complete the form below and we will contact you to arrange a demo of Luminance.
Please complete the form below to receive your Luminance brochure.
We will be in touch with you soon.