Contract: automated creationLegal expert systemLegislation: representation
- What does it claim to do?
- Substantiation of claims & potential issues
- How might the end-user assess effectiveness?
- What form does it take?
- Is it currently in use?
- The creators
What does it claim to do?
DataLex is an online application for creating further web-based mini-applications that can answer legal and compliance questions by means of a text-based question and answer dialogue. The system integrates with AustLII, the Australian database of legislative materials.
Claimed essential features
- Build internet applications for answering legal questions.
- Assemble legal documents from end-user inputs.
- Directly import information from public legal databases.
“AustLII’s DataLex inferencing software allows the development of Internet-based applications combining knowledge-based inferencing, a limited form of example-based inferencing, and automated document assembly.” (DataLex legal knowledge-base systems; archived)
“[P]rovides an easy-to-use environment in which end-users conduct a question-and-answer dialogue with the application in order to provide information (‘facts’) to it in order for the system to draw conclusions, and to conclude a user session by producing a report (and in some cases a document)” (DataLex legal knowledge-base systems; archived)
Claimed rationale and benefits
- Develop online legal reasoning applications.
- Assist legal advisory services, compliance, and decision support.
“DataLex can be used to develop legal reasoning applications in areas such as legal advisory services, regulatory compliance, decision support and Rules as Code.” (Home; archived)
Claimed design choices
- Integrates with public databases of legal sources: DataLex outputs can be linked to source texts.
- End-users can ask a DataLex app why it reached a particular conclusion.
- Uses yscript, a declarative programming language which is easily intelligible.
“integration between the inferencing components and the legal sources located on LIIs is one of the principal distinctive features of the DataLex approach” (DataLex legal knowledge-base systems; archived)
“DataLex Application Development Tools provide import, edit and pre-preprocessing functions which assist app creation and testing.” (Home; archived; the tool is available at http://datalex.org/dev/tools)
Allows users to ask “Why questions are being asked and How conclusions have been reaching, as well as to Forget facts previously provided, and to test hypothetical facts” (DataLex legal knowledge-base systems; archived)
“yscript uses a light-weight natural language parser to transform simple English language propositions into their positive or negative form and to turn them into questions when necessary” (Coding in yscript: A description of the yscript language, p. 40; archived)
Substantiation of claims & potential issues
- Depending on the quality of the translation of legal provisions into DataLex consultations, the output might not be legally sound (but note that the authors of DataLex explicitly mandate that it be used only for educational and testing purposes).
- Those coding DataLex interpretations of the law may not have the legal authority to conclusively determine the meaning of the rules (since they will usually not be working in a judicial capacity). This could be problematic if they are relied upon as authoritative statements of the law and are assumed to have legal effect.
- Maintaining DataLex representations of legal provisions may impose significant overhead due to changing judicial interpretations of (i) the original legal provisions, and (ii) fundamental rights that impact their application.
- There may be a risk that DataLex is used to formalise areas of law that are necessarily too complex or ambiguous for the kind of deductive reasoning the system is capable of.
DataLex has been in development since 1987 (Greenleaf et al., 1987) and is now hosted by the Australasian Legal Information Institute It should be regarded as a research project in the realm of Rules as Code.
- Datalex focuses on Australian, as well British and Irish, legislation.
- Examples of Datalex applications can be found at https://austlii.community/wiki/DataLex/. They include Australia’s Foreign Relations Act consultation, Non-Disclosure Agreement consultation and UK Occupation Order consultation, amoung others.
DataLex brings together several elements into a single product/platform aimed at formalising (Australian) legislative texts to enable various ‘consultations’ or ‘apps’ to be built. Although the DataLex application is publicly accessible through a web interface and the yscript language is well documented, the source code of the project is not publicly available. The extensive collection of (user) manuals therefore gives the most detailed view on the internals of DataLex.
The creators schematise the various elements as follows:
Figure 1. From the DataLex Developer’s Manual, p. 7
Red bubbles are external sources of canonical legal texts; green are the formalised ‘rulebases’ shared on the DataLex wiki in the yscript language; blue is a module for automating citations between documents; purple are the back-end components that execute rule-based reasoning; grey is the DataLex Application Development Tool where the user creates or edits a consultation (app). The latter brings it all together in a single interface for creating consultations. These are then executed by pointing the DataLex inference engine at a valid rulebase, which is usually stored in the wiki (e.g. http://datalex.org/app/?rulebase=http://austlii.community/foswiki/DataLex/NonDisclosureAgreementKB)
Core to the DataLex project is the yscript language: a rule based language in the declarative programming paradigm, that allows for small legally annotated decision scripts to be written. The syntax stays very close to the English legislative texted, but is extended with relevant keywords like ‘RULE’ to assist further computation and reasoning. The manual of the languages suggests that it resolves programs trough a system of forward an backward pattern matching, and user intervention, until no matching rules can be applied, or the users has provided all missing data. Combined with a basic type system, DataLex programs can guide users along the application of a rule or regulation. The type system allows rules to have threshold conditions for application such as an age limit.
The reasoning with yscript is done using the yscript-interpreter. The engine reasons with both forward-chaining (asking if questions to reach a fact conclusion) and backward-chaining (testing a fact conclusion by asking if questions).
Case-based reasoning: the system can also do some basic case-based reasoning, using ‘Precedent Analysis by Nearest Neighbour Discriminant Analysis’ (PANNDA) to compare bundles of facts to existing ‘examples’ that formalise existing cases into sets of conditions. However, one should not expect the engine to be performant for large rulesets, nor expect the reasoning to be applicable in situations where the rule-application is more advanced then pattern matching. (e.g. defeasible rules.)
While legislation is generally formalised by hand, the ylegis preprocessor can automate (at least initially) the conversion – even directly importing a legislative instrument from AustLII. The preprocessor is integrated into the DataLex Application Development Tool. The syntax of the yscript language allows for an annotation of the rules, and marking of their origin in the original legal text.
The DataLex Application Development Tool brings all the tools together, allowing for the development of yscript programs.
Rationale and benefits
- These programs are served to users as ‘consultations’. These consultations are available through in chat-dialog in a web interface. However, AustLII clearly warns its users that these applications are currently not suitable for any use outside of the experiment:
Figure 2. Disclaimer about DataLex’s use beyond education and testing
DataLex legal knowledge-base systems (archived) provides a solid overview of the system’s components and methodology.
Rule-based inferencing (I): Knowledge-bases and rules (archived)
Case Based Inferencing Using DataLex, inc PANNDA implementation (archived)
- G. Greenleaf et al, ‘Building datalex decision support systems a tutorial on rule-based reasoning in law’ 
- A. Mowbray et al, ‘AustLII’s DataLex Developers’ Manual’ (2021) (archived)
- A. Mowbray, ‘Coding with yscript – A description of the yscript language’ (2021) (archived)
- A. Mowbray et al, ‘The DataLex legislation preprocessor for rules as code’ (2021) (archived) – a short precis of the ylegis preprocessor.
- Greenleaf, Graham, Andrew Mowbray, and Alan L. Tyree. “Expert systems in law: The DataLex project.” Proceedings of the 1st international conference on Artificial intelligence and law. 1987.
- G. Greenleaf et al, ‘Representing and using legal knowledge in integrated decision support systems - DataLex WorkStations’ (1995) 3 Artificial Intelligence & Law 97-142 https://papers.ssrn.com/abstract_id=2183481
- One of the original theoretical and technical explanations of the approach – see Part 4 for a description of the inferencing engine
- A. Mowbray et al, ‘Utilising AI in the Legal Assistance Sector—Testing a Role for Legal Information Institutes’ (2020) 38 Computer Law & Security Review (sets out some technicalities of the rule-based reasoning approach on pp. 3-4)
- G. Greenleaf et al, ‘Building Sustainable Free Legal Advisory Systems: Experiences from the History of AI & Law’  University of New South Wales Faculty of Law Research Series
- A. Mowbray, P. Chung, and G. Greenleaf, ‘AustLII’s DataLex AI platform and its relevance to Law Faculties: The DataLex Development Environment’ (2021) Australasian Law Academics Association (ALAA) Conference, USyd and UTS
How might the end-user assess effectiveness?
Users can access the system to build an app ‘consultation’. Number examples are providedat http://austlii.community/wiki/DataLex, covering various legal instruments (mostly Australian but also UK). The ruleset can be compared with the originating legal instrument to ascertain the quality of formalisation.Top
What form does it take?
The project brings together various components into a web-based platform that facilitates the creation of ‘consultation apps’.
Rules can be executed as discrete ‘consultation apps’ for particular legal instruments, e.g. ‘Australia’s Foreign Relations Act’ (and see the ruleset that the web-based application ‘ingests’ to create the consultation)
There is no cost or subscription involved, however (see Q10 below).
Existing templates can be used to execute ‘consultations’ based on pre-formalised sets of rules, or alternatively new sets can be written for novel purposes/laws.Top
Is it in current use?
At time of writing it is not clear if DataLex is used in practice, however it appears the creators want it to be – the documentation is up-to-date (2021) and the website has recently been updated to look more commerce-friendly. All existing consultations have the disclaimer mentioned above, however.Top
Legal publisher, Academics
Developed by AustLII (the Australiasian Legal Information Institute), a joint operation of the University of Technology Sydney and the University of New South Wales. The main parties are:
Andrew “has degrees in Computing Science and Law. He wrote the sino search engine, the hypertext markup software, and the ysh inferencing engine … Andrew wrote the DataLex Workstation Software, the AIRS free text retrieval engine and the LES legal expert system shell.”
Philip Chung “holds degrees in Economics and Law from the University of Sydney, with honours in Computer Science and Operations Research as well as a PhD in Law from UNSW.”
Graham Greenleaf “has degrees in Arts and Law, and is a Fellow of the Australian Computer Society. He has researched and taught in the areas of intellectual property, privacy, cyberspace law and computerisation of law since 1982.”
The project was started in the 1980s and has gone through various phases (G. Greenleaf et al., ‘Building Sustainable Free Legal Advisory Systems: Experiences from the History of AI & Law’ (2018) 34(1) Computer Law & Security Review 324). The earliest paper on Google Scholar dates from 1985.Top
Background of developers
Potentially any English-language jurisdiction that has legislation amenable to representation in yscript, though DataLex has features for automatically importing legal text from AustLII, so to that extent it targets Australian jurisdictions.
Target legal domains
Any that is governed by legislative rules that lend themselves to a questionnaire format. The website specifically mentions that DataLex can be used “to develop legal reasoning applications in areas such as legal advisory services, regulatory compliance, decision support and Rules as Code” (Home; archived).Top
“we haven’t figured out what we are going to release the yscript under but we are going to release the actual interpreter so that the yscript language is available. Were still sorting through which of the licences we’re likely to use …” AUSCL Rules as Code Masterclass - AustLII DataLex video (11 May 2021) https://www.youtube.com/watch?v=8fTqQcVyVaI
“The DataLex Development Environment is freely available for non-commercial use and may be used by Australian and overseas law schools. Any individual users are welcome to use the DataLex tools and documentation to test and run their own apps for personal educational purposes. Government agencies, and private organisations (NGOs) with public interest objectives are welcome to contact AustLII to discuss possible use of the DataLex package. If you wish to discuss using DataLex for commercial projects, please contact AustLII’s Development Manager: Richard Hunter email@example.com.” (A. Mowbray, P. Chung, and G. Greenleaf, ‘AustLII’s DataLex AI platform and its relevance to Law Faculties: The DataLex Development Environment’ (2021) Australasian Law Academics Association (ALAA) Conference, USyd and UTS)Top