Della AI is a doc analysis company that has a strong focus on allowing lawyers to ask questions in natural language and get back a clear answer. The idea is that in real life lawyers don’t think about contracts in terms of ‘extracting clause X or Y’, but rather just want to find the answer to a query.
Below, Hugo Seymour, COO of the UK- and French-based company, shows how the system operates in this AL TV Product Walk Through.
This also covers areas such as raising a red flag on an aspect of a contract as part of the workflow, and its multi-lingual capability, which can cover a lot of languages at the same time. It can be used for single doc review as well as large-scale event-driven contract reviews.
(Press play to watch inside the page – approx. 12 mins.)
All well and good. But how does it do it? This is how the company explained things to this site:
‘Della uses cutting-edge massive language models to power its AI and an innovative system called transfer learning which allows clients to share training to deliver value faster. This reduces time-to-value for large organisations and puts sophisticated systems within the reach of smaller firms that would not have had the time to train their own. We think these network effects will allow us to solve document review.’
And what do they mean by a ‘massive language model’?
‘This is short for ‘massively pre-trained language model’, so any of the BERT/XLM/T5 variants fit the bill here. The distinction between a language model and technologies like word vectors is that the former have an idea of how a syntax and the grammar interact, whilst the latter only really look at the words in isolation. The ‘massive’ part comes from the size of the dataset that people use to build these.’
So, there you go. The legal AI world remains a rich environment with a wide variety of companies, with each trying to take the sector onwards. You can find out more about Della here.