Law as Code: Introducing AustLII’s DataLex AI
7 Pages Posted: 28 Nov 2021
Date Written: November 16, 2021
AustLII, as a provider of free access to legal information, has a distinctive approach to the use of artificial intelligence (AI) which is called ‘DataLex’. This article introduces that approach, and particularly some recent innovations.
The DataLex approach is based on symbolic representation and processing of information, particularly of sets of rules, a less common approach than the current preference for systems based on machine learning (ML). The DataLex approach aims to represent what rules say, not what they do, and it works within the constraint of assuming that resources available for AI developments will be limited.
The seven technical components contributing to the DataLex approach are introduced: (i) the yscript language for representing and manipulating propositions; (ii) the yscript interpreter for running yscript code; (iii) the ylegis pre-processor program which converts the text of legislation into yscript rules that reflect the legislation’s structure; (iv) the formal mode of ylegis which enables code to be written which is very close to the natural language (English) version of the legislation; (v) the DataLex application development tools, which anyone can use to develop test apps; (vi) the default user interface for consultations; and (vii) the DataLex Community development environment. The ylegis pre-processor is the most significant innovation, a promising approach to scaling up ‘law as code’, but only in the early stages of development.
The ‘DataLex approach’ is not technically determined by these tools, but is instead a set of recommendations for how to use them most effectively. These conceptual elements are explained. ‘Open access’ to elements of the DataLex toolkit is also an important part of the approach.
Keywords: law, rules as code, law as code, AI, AustLII
Suggested Citation: Suggested Citation