I'm new to AI development and am looking for a quality algorithm (potentially nlp?) implementation proved against US legal texts.

Obviously some training would need to be done, but I've found little to no online references to go on when it comes to running assessment against US legal documents.

My goal is to use an algorithm to discover potential issues in long and complex legal texts, or associated (groups) of legal texts which bind one or more related entities (people or corporations) to potentially conflicting clauses.

Just a pointer in some kind of direction would be helpful.

  • 2
    $\begingroup$ Welcome to AI.SE.....I am not an expert.....but legal issues always have a lot of context and human emotions attached to it.....and that is exactly where any nlp program or an expert linguist will fail to provide a satisfactory solution... $\endgroup$
    – user9947
    Mar 12, 2018 at 18:59
  • $\begingroup$ Context I think is absolutely true, but I'm less concerned with emotion. Lawyers are focused on getting the best result for -their- client. But the lawyers for the others involved may not be as diligent, or may be more diligent. The goal should be to highlight glaring deficiencies, loopholes or other likely issues for further human review. $\endgroup$ Mar 12, 2018 at 19:42
  • 1
    $\begingroup$ Have you seen the work done by Ross Intelligence using IBM's Watson? ibm.com/blogs/watson/2016/01/ross-and-watson-tackle-the-law $\endgroup$ Mar 13, 2018 at 3:50
  • $\begingroup$ @BrianO'Donnell I have seen that post previously, yes. While Watson may understand law to some extent, the solution is controlled by IBM and can change without my control. My goal is similar though: Translate legalese to common English and highlight problematic legal language for review. I anticipate running this effort in two areas: for-profit to review contracts and other legal documents; non-profit to review legislation for issues detrimental to the public-good. Control over the system is necessary to ensure the unencumbered operation of the platform. $\endgroup$ Mar 13, 2018 at 14:03


You must log in to answer this question.