deBessonet, C. G. and Cross, G. R., "An Artificial Intelligence Application in Law: CCLIPS, A Computer Program that Processes Legal Information," Berkeley High-Technology Law Journal, 1(2), 329-409, Fall 1986. |
| Abstract: The authors describe the current state of
an ongoing project, in which the project's purpose is create a computer system to process
legal information. The goal of the Civil Code Legal Information Processing System (CCLIPS)
is to analyze facts presented to it, retrieve relevant cases and statutes from Louisiana's
Civil Code and determine the legal effects of the facts. The authors first describe some
of the basic principles of artificial intelligence, next describe the basic features of
CCLIPS, then examine how artificial intelligence techniques can be used in the legal
domain, and conclude with the future direction of work on CCLIPS. In order for a system such as CCLIPS to work, information needs to be made available to the computer in a form that it can understand. Knowledge representation is therefore an important aspect of the system. Ideally, the system would be able to do natural language processing, where the computer can analyze oral or written text and create its own representation. However, computers are not yet able to do this in a generalized manner. For a conceptual retrieval system to work, the computer must be able to retrieve not only important facts, but also higher-level significant themes. For statutory interpretation, this process is aided by the inherent categorization of the statutes. However, the statutory construction still needs to be hand-crafted at this time. CCLIPS currently uses a structured language called Atomically Normalized Form (ANF) to represent information. It explicitly models states of the world, events, relations of different types, and beliefs. Causality is one of the most important real-world phenomenon, and the system uses causal structures to create deep models of knowledge representation. Temporal representation is also an important part of the reasoning process, and needs to be represented explicitly. The authors present examples of building ANF models. They conclude by anticipating that a system like CCLIPS will eventually be used not only for statutory retrieval but also statutory drafting. |
