Скачать 262.02 Kb.
Representing and using legal knowledge 4
Representing and using legal knowledge in integrated decision support systems: DataLex WorkStations
Graham Greenleaf, Andrew Mowbray and Peter van Dijk
Published in Artificial Intelligence and Law Kluwer, Vol 3, Nos 1-2, 1995, 97-124; Published on the web as part of the authors' web pages.
Associate Professor of Law, University of New South Wales Sydney 2052 Australia; firstname.lastname@example.org
Senior Lecturer in Law and Director of Undergraduate Programmes, University of Technology, Sydney, Broadway 2007; email@example.com
Peter van Dijk
Lecturer in Law (P/T), University of Technology, Sydney, Broadway 2007
Abstract: There is more to legal knowledge representation than knowledge-bases. It is valuable to look at legal knowledge representation and its implementation across the entire domain of 'computerisation of law', rather than focussing on sub-domains such as 'legal expert systems'. The DataLex WorkStation software and applications developed using it are used to provide examples. Effective integration of inferencing, hypertext and text retrieval can overcome some of the limitations of these current paradigms of legal computerisation which are apparent when they are used on a 'stand-alone' basis. Effective integration of inferencing systems is facilitated by use of a (quasi) natural language knowledge representation, and the benefits of isomorphism are enhanced. These advantages of integration apply to all forms of inferencing, including document generation and case-based inferencing. Some principles for development of integrated legal decision support systems are proposed.
Key words: integrated, natural language, hypertext, document assembly
1. 'Integrated' computerisation of law
1.1. Meanings of 'integration'
1.2. DataLex WorkStations
2. Hypertext and text retrieval engines
2.1. The hypertext engine
2.2. The text retrieval engine
3. Integrating hypertext and text retrieval
3.1. From searching to browsing
3.2. Browsing to searching
3.3. Searching for smarter searches
4. The inference engine
4.1. Rule-based inferencing
4.2. The knowledge representation: (quasi-) Natural language
5. Integrating inferencing with hypertext and text retrieval
5.1. The limits of inferencing
5.2. Explanation facilities in the Workstation
5.3. Distributing access to the knowledge base
5.4. User Control over inferencing - a reason for integration
5.5. Isomorphism and The value of 'English-like' representations
5.6. Isomorphism and Developer control of inferencing
6. Multi-modal inferencing and integration
6.1. Document Generation and the inference engine
6.2. 'Examples' and the inference engine
7. Future work on integration
7.1. Exploiting common elements in different representations
7.2. Terminology and theory for integration
1. 'Integrated' computerisation of law
There is more to legal knowledge representation than knowledge-bases, just as there is obviously more to the computerisation of law than legal inferencing systems. It is valuable to look at legal knowledge representation and its implementation across the entire domain of 'computerisation of law', rather than focussing on sub-domains such as 'legal expert systems'.
[Paquin et al 1991] take a similar approach:
We think that a shift from a dominant mono-technological point of view, that is, a legal expert system, to an integrated point of view where the user is included in the same way as several technologies, is more productive in terms of a real world scale system. This point of view relinquishes the pretensions of automated legal decision-making. It considers those systems for what they are: information systems that help human decision-making.
The perspective from which our work derives is that of a developer of legal computer applications intended to be of immediate practical utility. One emphasis of our work is therefore on how to make the best use of current technologies. From this perspective, we first argue that effective integration can overcome some of the limitations of the current paradigms of legal computerisation which are apparent when they are used on a 'stand-alone' basis. The use of integrated tools in the building of ‘real world’ applications can save application developers from attempting to use the techniques of one mode of computerisation for purposes for which it is not suited. It is sometimes difficult to anticipate at the outset of a project what combination of tools will be needed. Attempting to force square pegs into round holes is rarely satisfactory.
Second, we argue that effective integration of inferencing systems is facilitated by use of a (quasi) natural language knowledge representation, and that this applies to all forms of inferencing. The advantages of such an approach for isomorphic knowledge representation is argued. Third, we argue that a theoretical foundation for development of integrated legal decision support systems needs to be developed, and we propose some basic principles for discussion.
In the course of these arguments we discuss ways in which legal domain knowledge can be represented by application developers for use by the text retrieval and hypertext components of systems.
Paradigms of computerisation of law
Computerisation of law has developed from a number of originally unrelated technologies: the development of online free text retrieval systems from the 1960s; the revival of artificial intelligence research in the form of expert systems in the 1970s, the development of automated document generators, and the ‘rediscovery’ of hypertext in the late 1980s1. Lawyers are interested in the computerisation of a number of different aspects of legal practice, including retrieval of documents relevant to decision-making, other forms of research, the decision-making itself, the generation of legal documents, litigation support and computer-assisted learning. We use ‘computerisation of law’ to encompass both the computerisation of these various aspects of legal practice, and of the legal source materials (such as cases, statutes, commentary) used in them.
Most commercially published legal applications have implemented only one of the four dominant paradigms. Text retrieval systems have, of course, been overwhelmingly dominant. Legal hypertext systems have become increasingly common in the last few years, and have usually been integrated with text retrieved but often only to a limited extent. There are still very few legal inferencing systems in commercial use, and their integration with hypertext or text retrieval is still exceptional. Automated document generators have been commercially available for over a decade, but only on a relatively small scale and with virtually no integration with other technologies. This lack of integration is not peculiar to law, but has been observed to be a general feature of the computerisation of information [Frei (1990)]. There are an increasing number of exceptions in commercial computerised legal systems, some of which are mentioned in this paper.
Each of the current paradigms have prompted theoretical legal and computing research, ranging from considerable research on such matters as the jurisprudential models implicit in various types of legal expert systems, and the adequacy of Boolean retrieval for legal research, to a far more limited amount of research on the principles of hypertext systems suited to law or legal document generators. Research on the relationships between these different approaches has been more sporadic., but includes substantial research on the use of AI techniques in information retrieval, and work on the development of integrated computerised workstations for public administration2. In the last few years it has become increasingly common for articles on legal computerisation to recognise the value of using hypertext in conjunction with other paradigms of legal computerisation. Vandenberghe stressed the importance of integration over a decade ago ([Vandenberghe, 1982]; see also [Koers, 1990]), and others have occasionally done so since [eg [Oskamp and van der Berg, 1990], [Paquin et al, 1991], [Greenleaf, Mowbray and Tyree, 1991], [Mead and Johnson, 1991], [Soper and Bench-Capon, 1994]). However, it is still true to say that there has been relatively little development of any general theory exploring a comprehensive approach to the integrated computerisation of law.
Limits of the paradigms as 'stand-alone' technologies
One of the main justifications for exploring the integration of these technologies is that each has well-recognised limitations and deficiencies when it is used on a 'stand-alone' basis for the computerisation of law. We will mention a few. Most users have significant problems in formulating effective queries in text retrieval systems (at least those based on boolean connectors and proximity), particularly where any legal concepts must be represented by the user in queries. With text retrieval alone, it is often difficult to go to neighbouring text items which are not in the retrieved list, in order to understand the context of a retrieval item. Hypertext systems require that all associations between text items be identified in advance by the system developer, whereas text retrieval systems at least allow some users to retrieve items which have unusual and unanticipated connections. Inferencing systems tend to be very brittle because of the endemic open textured nature of legal language, and it is difficult to build and maintain adequate explanatory mechanisms which assist the user to resolve these problems, or explain system outcomes. Differences between each of the paradigm technologies in terms of noise and silence (relevance and recall), and in costs of development in relation to coverage, are suggested in [Paquin et al, 1991] at 2.2.
|PhD: University of Sarajevo, Sarajevo, Bosnia-Herzegovina, June 1, 1991 Thesis: “Systems Approach to Modeling of Integrated Decision Support Systems” msc: University of Sarajevo, Sarajevo, Bosnia-Herzegovina, June 27, 1987||Decision Support and Expert Systems (mis 538)|
|Keywords Distributed decision, constraint propagation, negotiation, decision support. Introduction||Information Systems Journals: Knowledge Castles or Knowledge Gardens? Brian Whitworth|
|The move to "open systems" requires re-thinking interaction with knowledge representation systems||Rdbm versus oodbm in Support of Integrated Data Bases|
|Chapter 9 e-business Decision Support Learning Objectives||Enabling Decision Support and Costing of Product Designs by using Visual Metaphors|
|A spatial decision support system for bank location: a case study||Manufacturing/Integrated Systems Technology Program|