Law as Code: Introducing AustLII’s DataLex AI

tags: DataLex

Notes

ML has had limited success in relation to legal information.

NOTER_PAGE: (2 . 0.5486526946107785)

[citation needed]

immediate conceptual fork between what legislative rules ‘say’ and what they are taken to ‘do’

NOTER_PAGE: (2 . 0.6302395209580839)

We regard legislative rules as a hierarchical set of propositions which include both real world references as well as propositions about how legislation works.

NOTER_PAGE: (2 . 0.6646706586826348)

extremely opinionated, which is fine, but I don't think I agree with either idea: that rules are strictly hierarchical or that they are strictly propositional

ask how can the creation and maintenance of legal code-bases be made simple and affordable enough so that the legal assistance sector (and not only the commercial sector) can be involved in it.

NOTER_PAGE: (3 . 0.09655688622754492)

DataLex approach therefore involves the use of a less common approach to AI (symbolic representation)

NOTER_PAGE: (3 . 0.19535928143712575)

stubbornly clinging to the "AI" label as its popular meaning drifts to the statistical. I respect it :)

The yscript language – A language for representing and manipulating propositions

NOTER_PAGE: (3 . 0.3532934131736527)

Code-bases of legislation cannot resolve the correct interpretations of the open- textured terms in the legislation. It is necessary for them to provide access to the source documents

NOTER_PAGE: (5 . 0.8226047904191617)