Suggestions
Graeme J.
once a lawyer... | now doing software stuff
Graeme J., an accomplished legal professional, holds an impressive educational background with degrees from prestigious institutions like University of Cambridge, University of Oxford, and University of Virginia School of Law.
With a wealth of experience in the legal field, Graeme J. has held various key positions in renowned organizations. He has served as the CEO at Juralio Ltd and also worked as a Consultant, Partner, Solicitor, and Trainee Solicitor at Herbert Smith Freehills.
Highlights
I've been doing quite a lot of informal experimentation over the last year or so using LLMs to test the design of the noslegal open source taxonomy.
This is a taxonomy intended to help with various practical legal needs but not designed for technical legal research (as that tends to occur in a highly technical conceptual context).
I've found LLMs quite useful in some respects. The question occasionally comes up in my mind, and in discussion, whether such models will over time render taxonomies irrelevant.
My intuition is 'no', as an effective taxonomy should draw valuable distinctions in a particular domain and value depends on things like
-
really understanding the relevant needs across quite a complex set of sub-domains and figuring out which are common, and which distinct,
-
drawing distinctions which are as clear and simple, as well as useful, as possible, and layering them in a 'just enough' way.
That's all quite difficult and requires, I think, some human discernment. I can't see inherently how it can be 'solved' by LLMs.
In fact, my intuition is that a properly human-designed taxonomy will be all the more valuable in future, though with LLMs being used to improve and test it incrementally, and to do things at various levels which don't actually need a specialist taxonomy.
I do try to keep an open mind on this though and one thing I've been keeping an eye out for is relevant research.
This June 2024 paper - "Are Large Language Models a Good Replacement of Taxonomies?" - by Yushi Sun and seven colleagues is something I found this evening on point, and it's an interesting read.
The most important conclusion, to my mind, is the one I've put in the image.
Thoughts very welcome.
https://lnkd.in/eqEFG62J
This post is to publicise a noslegal event being held in London on 2nd October 2024. It may be of interest to people who work with data at law firms and in-house legal, addressing practical questions such as those in the image.
noslegal is a voluntary community focused on developing taxonomy and other data standards for legal work, holding related discussions and releasing material on open source terms (i.e. anyone can do anything with them).
Website - https://www.noslegal.org/
We are at a very interesting stage now, with different elements of the work at differing stages of maturity. There will be discussion over the next few months on next steps, with a plan being agreed in October.
One of noslegal's sponsors, iManage, is kindly hosting this event at their EMEA headquarters in Shoreditch.
It will start at 2pm. There will be a few hours of substantive discussion then a chance to chat and meet people informally over snacks and drinks.
We have a lot of participants already from large and medium UK law firms and from some software companies. But others doing relevant work (for example, people in legal ops or similar roles in legal departments) are also very welcome.
Eric Eck has also recently started to increase noslegal's visibility in continental Europe. We haven't yet organised an event there, but you're welcome to come in to London for this one!
Please DM me or Eric if you're interested and we can tell you more. We do have some limits on numbers but will try to do what we can.
Many thanks to iManage for making this event possible Robert Florendine Sam Grange Jack Shepherd
