LLD(ID:5643/lld001)KR Language for lawfor Language for Legal Discourse McCarty, Rutgers, 1989 AI language for recording legal knowledge, used to write TAXMAN II Places Related languages
References: in Campbell, C. editor, Data Processing and the Law, (Sweet and Maxwell, 1984 view details in Campbell, C. editor, Data Processing and the Law, (Sweet and Maxwell, 1984 view details in Fiedler, H. ; F. Raft, and R. Traunmilller, editors, Expert Systems in Law: Impacts on Legal Theory and Computer Law, Attempto-Verlag, Tiibingen, 1988 view details in Fiedler, H. ; F. Raft, and R. Traunmilller, editors, Expert Systems in Law: Impacts on Legal Theory and Computer Law, Attempto-Verlag, Tiibingen, 1988 view details in Fiedler, H. ; F. Raft, and R. Traunmilller, editors, Expert Systems in Law: Impacts on Legal Theory and Computer Law, Attempto-Verlag, Tiibingen, 1988 view details in Fiedler, H. ; F. Raft, and R. Traunmilller, editors, Expert Systems in Law: Impacts on Legal Theory and Computer Law, Attempto-Verlag, Tiibingen, 1988 view details In two previous papers on the prospects for intelligent legal information systems, I advocated the development of "deep conceptual models" of particular legal domains. My motivation was both practical and theoretical. On the practical side, I argued, our long-term goal should be an integrated analysis/planning/retrieval system that matches as closely as possible the way a lawyer actually thinks about a legal problem. On the theoretical side, my work with Sridharan on the TAXMAN project had clarified the importance of an adequate domain theory in any attempt to model the arguments of lawyers in hard cases, For both purposes, I claimed, deep conceptual models are essential. Although some commentators have expressed puzzlement about the meaning of the term "deep conceptual model", the basic idea is easy to state. There are many common sense categories underlying the representation of a legal problem domain: space, time, mass, action, permission, obligation, causation, purpose, intention, knowledge, belief, and so on. The idea is to select M small set of these common sense categories, the ones that are most appropriate for a particular legal application, and then develop a knowledge representation language that faithfully mirrors the structure of this set. The language should be formal: it should have a compositional syntax, a precise semantics and a well-defined inference mechanism. The semantic interpretation of the common sense categories should be intuitively correct, that is, it should generate exactly those entailments that ordinary people (and ordinary lawyers!) generate in similar situations. The inference mechanism for the language should be complete and sound, in principle, but, in practice, completeness and soundness would often be sacrificed for computational tractability, just as they are in ordinary human (and ordinary legal!) reasoning. Clearly, if a language of this sort could be developed, it would provide a uniform framework for the construction of a legal analysis/planning/retrieval system, and a solid foundation for further theoretical work. In this paper, I will describe the basic features of a Language for Legal Discourse (LLD), which takes a first concrete step towards the realization of this goal. An example of the surface syntax of LCD is shown in Figure 1. This example is from the TAXMAN II project, and it shows the influence of the frame-based representation language AIMDS in which TAXMAN II was originally implemented. The difference lies in the fact that each syntactic feature of LLD has a precisely defined semantic interpretation and an associated in- ference procedure, which was not the case for the AI representation languages of the late 1970’s. The design of LLD has also been influenced by the work of Bonner and McGuinness on a seller’s remedies for breach of contract under Article 2 of the Uniform Commercial Code, and by my joint work with Dean Schlobohm on estate planning with prototypes. As of this writing, the language has been partially implemented (in Com- mon LISP on a SUN/3 workstation): the various in- ference mechanisms have been specified, a parser from the surface syntax of Figure 1 to the compiled proof procedures has been written, and a unification algo- rithm that handles sorted count terms and mass terms (see Sections 2.2 and 2.3) has been thoroughly tested. However, most of the work on LLD to date has been devoted to the theoreticsl foundations of the language. My plan is to write several papers on a Language for Legal Discourse, of which this paper is the first. Section 2 explains some of the peculiar features of the atomic formulae in LLD, and Section 3 describes the rules and proofs for the first-order sub-language, which is intuitionistic rather than classical. Section 4 then explains how these rules and proofs are extended to handle various modal features, such 8s time, action, permission and obligation. One important issue in the design of LLD concerns the integration of these several features into a single language, and this issue is addressed in Section 5. My point here is that the intuitionistic semantics at the core of the language facilitates the development of an integrated system, and simplifies the proof procedures for the various modalities. DOI in Proceedings of the second international conference on Artificial intelligence and law view details in Proceedings of the second international conference on Artificial intelligence and law view details |