The world of NLP-based tools in the legal sector, broadly termed ‘legal AI’, has evolved rapidly. Artificial Lawyer therefore has put together a new map of where we are now.
Back in 2016 when this site first made a map of Natural Language Processing legal applications there were just a handful of branches. In fact, back then a lot of the emphasis was just on trying to get across what NLP was and how it fitted into the wider world of ‘AI’ software. Now it’s fair to say it is truly an entire field of its own.
The table below, which is an Artificial Lawyer and TromansConsulting production, could be even more detailed, with branches separating out in more and more sub-branches. However, this hopefully gets the point across that NLP tech in legal has both proliferated across the market and diversified into multiple use cases.
A Guide to the Map
As you can see, the starting point is NLP, because for this site that is the root of this segment of the legal tech market. Take the NLP out and there is not much left of these applications. Also essential to these types of software tool is machine learning, whether that has been ‘baked in’ before someone uses them, e.g. as you might see in a legal research tool, or where there may be a mix of pre-set NLP approaches with the ability to add your own training to improve results, as we see in the field of post-execution doc review and analysis tools.
There is also a small outlier group of NLG, or Natural Language Generation tools, in the legal tech world. These remain a relatively niche area.
- Pre-execution – what we could see as the start of the journey, i.e. that phase where contracts are created, negotiated, red-lined and edited before signature. We see a growing range of tools here helping with risk identification, to automated markup, to playbook comparison and preferred text editor systems.
- KM – here the central focus is around using NLP to really get into the docs at a level of detail that reveals far more useful information than just the doc title, or what a key word search would provide.
- Business Info or (BI) – another growing area, which branches first into using doc analysis tools to capture and reveal contract data, often this is used in an integrated way with a CLM platform (this could also be added to the Post-Execution family, but it has been put here as the main focus is often on info of value to the wider business). The second branch is using NLP to study text narratives in bills to better understand if a law firm’s invoice matches the billing guidelines of the client.
- Post-execution – the one area that ‘legal AI’ is probably best known for, i.e. large-scale review of contracts, often for event-driven projects, e.g. due diligence in M&A. This has really branched out now, covering a wide range of transactional work (and not all of it is listed here), to repapering projects, to compliance, and of course is part of the eDiscovery field from building rapid taxonomies of what is in a doc stack, to helping with the review process (albeit used with a different outcome in mind from transactional review).
- Litigation Analytics – within this is placed the huge field of case law/litigation research, along with more specific use cases around participant analysis, e.g. looking at what motions a judge may have permitted in the past. To this we could also add ‘litigation prediction’, although in many ways this is just a reworking of the data that is gathered across this branch. And then we also have new areas such as brief drafting, where NLP explores a corpus of case law to reveal related material that may go into a brief.
- NLG – as mentioned this is where the software generates new text. One example is in patent application generation, where a large quantity of text is analysed and new text for an application is generated. This remains a fairly niche area in the legal tech world, for now, but could see rapid growth as new language generation models arrive.
- ‘Other’ – and there are other sub-branches we could add across this map to refine it even more, but for the space we have this hopefully covers many of the key aspects.
NLP tools in the legal world have evolved, ‘legal AI’ now is a broad family of use cases far beyond M&A due diligence review, although that remains a key use case. We can also expect this change and diversification to continue. New methodologies may arrive and no doubt NLP software will improve. E.g. we still don’t know for sure yet what impact the work on new language models such as GPT-3 will provide. There are also a lot of interesting outcomes that may occur by connecting the above branches with other applications, and we are already seeing some, e.g. NLP + CLM, or NLP + billing systems.
If we can get to such a diverse and multifunctional ecosystem of NLP uses in just a few years, then it would seem likely that there will be a lot more invention and iteration to come. This site is certainly looking forward to that.
(Note: please don’t cut and paste the map or use without permission, thanks.)