Our intern Margaux McQuilton, a student from Jersey's Institute of Law, has been busy over the last 4 weeks working on real & imaginary provisions of legislation for us.
First she parses them - looking for “if”-“then” structures, “and”, “or” & “not” provisions, and “must”/“must not” (to which she adds “if the person does [what the person must not do] then the person counts as in breach of this provision”). She then marks them up - with coloured highlighting so far, maybe XML later.
Then she turns that into questions & conclusions, sticking to the wording of the legislation.
Then she turns those questions & conclusions into markdown to generate flowcharts on Mermaid's free website https://mermaid.live.
Then she writes them for QnA Markup https://qnamarkup.org, a free web-based questionnaire system from David Colarusso at Suffolk Law School's Legal Innovation & Technology Lab.
In all this she is still sticking to the wording of the legislation. That means all we are capturing is the logical structure of the legislation, not the meaning of undefined expressions, and we are not adding any additional interpretation.
Next steps will include trying to get generative AI (ChatGPT, Bing Chat) to write the QnA & Mermaid elements, and looking at LegalRuleML & other markup for tagging the text (preferably in a way that a free reasoning program can work with). We will also look at whether we can publish the whole set of documents once they are finished.
Fantastic to see this explanation. Has the work so far generated tips for how drafters can write legislation that is easier to express as code?
Great work! Really interesting to read about the process and the initial steps.