Ask YC: AI/NLP for legal analysis?

I believe much of the legal texts in the US, and other legal systems, can be parsed into RDF triples/tuples/facts [0].

Of course people are trying to do this for all kinds of texts, but legal texts have a very specific structure, and a smaller vocabulary, which may make them easier to parse them.

Once we have a set of tuples it would be very easy to build an expert system that could replace a lawyer [1].

The system does not need to be 100% accurate. Each fact would keep a link to the original paragraphs in the corresponding legislature. This will allow people to improve the knowledge base of the system, wiki style. Also, if the system fails to parse a particular paragraph, but can understand the subject, for example, it can still present the particular paragraph when an inquiry is made about the subject. The user will then read the paragraph and enter its contents in tuple form.

Cases that form precedents can be a caveat. They usually are less structured than legislature, so it can

Has anyone tried something like this before? It definitely seems like a market begging to be shrunk.

[0] RDF triples are usually (subject, predicate, object). However, for legal texts (subject, predicate, object, condition) might work better.

[1] Of course, lawyers do a lot of other things, but for the sake of this idea let’s consider lawyers are paid $400/hr just to tell you what the law says.