Eigen Technologies, the legal and financial services AI doc analysis company, has managed to increase its data extraction speeds by between two and five times on average, they claim. This is seen as a major step forward for the analysis of very large contracts, which are often seen in the financial services sector.
The UK-based company said that the changes, which are part of its new 3.0 software release, mean that it can answer 10 specific questions about a highly complex contract of 500 pages in around five to seven minutes. Smaller, more standard, contracts would also therefore see questions answered far faster as well.
Eigen allows users to ask specific questions about a document stack, then receive answers directly via the NLP system.
This capability is especially valuable when a client is seeking to review a large number of massive contracts, which if handled purely on a manual basis would take many times longer.
That said, there is no ‘industry standard’ or ‘normal benchmark’ as yet for the legal market in terms of contract review speed. In part this is because each legal AI company is using their own blend of NLP and other text or pattern analysis approaches to do their work, so they are hard to compare directly in that way. Also, the exact way that each system interacts with a user, e.g. answering a natural language question, or perhaps just extracting and presenting sections of text for a certain type of clause, also vary considerably.
Eigen 3.0 has also a completely redesigned user interface, which they have found helps to speed up NLP training by 30%, the company said. It also includes features that enable improved handling of documents inside the platform, they said.
Chief Product Officer of Eigen Technologies, Ashley Fidler, said of the new update: ‘Our customers asked us for three things in Eigen 3.0: a better super-user workflow, improved performance for large teams working simultaneously, and easier document handling. We will be building on this in 2020 and will have further updates to come.’
1 Trackback / Pingback