Law (judiciary) contains such a huge corpus to apply NLP to, but yet there are only search engines designed for Law. Why is NLP not yet extensively applied? Is it because of dimensionality?
closed as too broad by Dan Scally, Peter, georg-un, timleathart, Stephen Rauch♦ Aug 21 at 2:08
Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. Avoid asking multiple distinct questions at once. See the How to Ask page for help clarifying this question. If this question can be reworded to fit the rules in the help center, please edit the question.
Welcome to the site and thanks for the great question! I recently led an NLP project that dealt with a lot of laws. While I have to obfuscate my actual work, here's a general view:
- The laws themselves may not be the best source data. It would take a massively transformed recordset in order to make most laws actionable for modeling. I'm talking about big rooms, full of lawyers providing an annotated version of laws in order to create a recordset that can actually be useful
- The above assumes that the laws have been digitized in some easy to digest format. That may not always be the case. In a lot of instances, you are referring back to classic OCR approaches as part of your data prep and I don't know anyone that likes working with OCR :-)
- The human-in-the-loop requirements are very high. So you have an algorithm, now what? That's not something you can just put out on Mechanical Turk for the layman to verify. You need more lawyers to help with the verification of your approach and correct mistakes that are happening
- Finally, you must get very sophisticated with your embedding layers in how you create and apply them. That's not an easy thing to do and very processor intensive - a GPU is highly recommended and not a lot of grassroot efforts are going to have this processing power
A friend (Law and CS graduate) recently wrote his PhD dissertation about the use of AI and ML approaches in law.
His conclusion is as you suggested (dimensionality), semantics, cultural concepts of justice, as well as non-binary data types (e.g. confessed but with what constraints and conditions?) will not lead to satisfying mathematical results, when it comes to the actual range of sentences.
However, NLP can help preprocessing cases to better distribute them to the person in charge of. In categorical cases (e.g. traffic delinquency) one can use NLP approaches to make the whole process more efficient.
NLP is very widely used in certain aspects of Law. I worked on few use cases related to Contract Management. While I can't talk about specifics, general areas where NLP is applied are :
- Distance analysis for paragraphs / sections of contract (v/s corpus of historical judgements)
- Automation of manual reviews and validations
- Automation of Business processes related to discovery
- NER (Specific to Legal domain)
Some articles on this topic :