I am responding to David Badertscher's informal survey about the future
of CALR. I have been working on developing artificial intelligence
approaches to CALR since the mid-1970's. Right now I am on sabbatical
at Harvard Law School, working on several related research projects.
Marc Lauritsen forwarded the message to me, and I am looking forward
to joining this discussion group.
One of my major research goals while at Harvard Law School is
to analyze, from an AI perspective, the intellectual
structure of current legal research aids (both automated and printed)
and re-think this entire structure from the standpoint of what we now
can do with knowledge-based software (e.g. rule-based expert systems
and frame-based conceptual indices). Below is an excerpt from a
recent grant proposal that explains some of my ideas more specifically.
Carole Hafner
Associate Prof., Computer Science
College of Computer Science
Northeastern University
Boston, MA 02115
tel. (617)-437-5116
tel. (on sabbatical at HLS): (617) 496-5355
Internet: hafner@corwin.ccs.northeastern.edu
A. Motivation
The legal profession, more than any other, is dependent on finding relevant
authority in published documents. Current on-line legal databases provide
access to more than 15 million documents; however, the difficulty of using
traditional full-text search methods to find relevant items is well
documented [Dabney 1986]. While numerous published indexes, digests,
treatises, and annotations exist in law libraries to assist legal researcher
this "engineered" knowledge is not brought to bear in computer-based
research tools. And in spite of the array of existing reference material, legal
research is perceived by many experts to be an overly complex and
inefficient process that needs significant improvements.
Artificial intelligence (AI) techniques such as expert systems and
computational linguistics offer potential for developing computer-assisted
legal research (CALR) systems that would be vastly superior to current
research tools. Expert systems can deliver the "engineered" knowledge
available in research handbooks, digests, etc. more efficiently than printed
books -- combined with hypertext, such systems could be revolutionary in
their impact. Full-text search would still be available in an expert system
environment, but it would be one of many options, not (as today) the only
automated tool. At the same time, AI offers new possibilities for legal
indexing based on structured concept representations such as semantic
networks and frames, which need to be explored.
As a first step in developing this new technology, we need to consider what
kinds of tools legal experts would find most useful and what kinds of user
interfaces would be most effective. Investigating this question is one of the
major research goals of this project. We will do this by bringing together
legal experts and computer experts in a 4-month seminar taught jointly by
Profs. Hafner and Martin.
Since engineered knowledge will always have gaps, it is also important to
develop more effective methods for full text searching. Improved methods
for text searching have wide applications not only in legal research, but also
in litigation support and office automation generally. Recent advances in
computational linguistics offer potential for improved text analysis and
retrieval systems; however, such methods need to be implemented and
evaluated to see whether this potential can be realized. Prof. Hafner has for
the past two years conducted research sponsored by the National Science
Foundation on improved methods for text analysis. During her one-year
sabbatical at Harvard Law School, she has support from NCAIR, the National
Center for Automated Information Retrieval, to implement and evaluate
these methods in a legal setting. This will be done in co-operation with Mr.
Lauritsen, whose previous research is in the area of automated legal practice
systems.
B. Description of Research Projects and Deliverables
The proposed research will focus on two specific activities: 1) designing an
"Expert Legal Research Workstation" and 2) developing linguistically-
oriented text analysis methods within a traditional full-text search
environment. Each of these activities is described below, along with its
tangible products.
B.1 Design of An Expert Workstation for Legal Research
This research will be done by Profs. Hafner and Martin and the students in a
seminar to be offered in the spring semester of 1992. The objectives of the
project are:
1) to advance our understanding of the mental processes involved in legal
research
2) to expand our vision of the potential of automated information systems in
law
3) to build a working prototype that incorporates these insights
The participants will analyze the process of legal research from an artificial
intelligence perspective, and develop a model of the goals, strategies, and
analytical models of both novice and expert legal researchers. We will use
this model to consider how electronically stored information can be
organized, searched, and presented in order to better support the legal
researcher. We are especially interested in discovering expert research
strategies, and incorporating them into our system design so that less expert
researchers can benefit from them.
A major activity of the seminar is a collaborative effort to produce our own
design for a legal researcher's "expert workstation", which will demonstrate
both the functionality and user interface that advanced CALR systems should
offer. Participants will each select a "tool" that they are interested in
developing and create one or several proposed designs to present to the
group. After discussion, we will settle on a preliminary set of tools to be
included in the prototype. Designs for these tools will be developed in detail
and computer prototypes created. Demonstrations will be given to legal
researchers in order to get their feedback; researchers in a variety of
settings will be included - courts, private law firms, government agencies,
etc., in addition to Harvard faculty and students. The prototypes produced,
while unlikely to be a final product that can be directly used by
professionals, are intended to provide a platform for future development of
useful research tools.
In order to make the expert workstation design appealing to potential users,
it must include a state-of-the-art user interface. Graphical user interface
software, along with prototyping tools offering hyptertext support, will be
required if the goals of the project are to be fully met. Innovative design
concepts will be explored to handle the unique aspects of legal research, such
as the need for exact citations of authority. The workstation design will
include these elements:
1. A variety of "tools" are activated in windows on the workstation screen.
2. Graphical user interface methods (i.e., icons and pointing devices) are
employed to make the tools easy to use and understand.
3. Researchers (i.e. users) can select tools according to personal preferences.
4. Tools can be customized to the individual user.
5. Traditional tools and knowledge based tools are both available.
6. Where possible, intelligent options will be added to traditional tools.
The following are some preliminary ideas about the kinds of tools to be
included in the workstation design. The primary objective of the project is to
evaluate these and other possibilities in light of the needs and stratgies of
people engaged in legal research on a regular basis, and to use that feedback
to guide the workstation development effort.
1. Citation tools. Once a potentially useful document is located, citation
tools allow the researcher to retrieve earlier documents which it cited or
later documents which cited it. Intelligent versions of citation tools can be
created which only link to documents which have a specified relationship to the
source document.
2. Text search tools. Researchers can provide descriptions of topics
or issues on which they want to find authority, and the computer will
directly analyze the text of documents. See Section B.2 below for a
discussion of intelligent extensions of traditional Boolean text search tools.
3. Topical index tools. Using a predefined topic index is currently one of the
most effective ways of finding relevant material using manual research tools.
Topic indices will be one of the most important components of the Expert
Legal Research Workstation.
4. Database tools allow traditional database attributes to be assigned to legal
documents, such as date, jurisdiction, court, etc. It is often useful to be
able to subset a document collection on these types of attributes.
5. Rule-based tools will provide expert advice on a number of problems,
including expert advice on finding sources of legal authority for particular
kinds of research problems.
6. Document structuring tools. Segmenting documents into different sections
promotes more efficient searching and browsing. In addition, the user may
want to attach documents together in a hypertext structure (for example, the
user can make notes and attach them to documents or link documents
together for future reference.
This archive was generated by hypermail 2b29 : 03/07/00 PST