Hypertext as a tool for information Gardening for legal Applications

AutoreMarc Nanard/Jocelyne Nanard
Pagine47-77

Page 47

@1. Assisting Information Gardening

In the legal domain, activities deal with huge and growing amounts of information which are mainly handled as documents. But unlike technical documents, legal documents contain subtle, informal, non-explicit aspects of information that cannot be neglected. So many legal tasks are still difficult to be computer-aided.

Currently a lot of techniques are used for legal applications, especially in fields where a preliminary modeling of information is possible. On the one hand, databases and information retrieval are concerned with storing and retrieving large amounts of legal documents. For instance, automatic transformation of SGML documents [Wilson 1990] as well as indexing and clustering techniques are used to produce hypertext databases allowing querying as well as browsing. On the other hand, a lot of works take advantage of A.I, techniques for building specific legal expert systems and take-advantage of computer assisted reasoning. All of these techniques suppose that a sound knowledge acquisition process has been achieved in order to build a domain model and an application model [Breuker 1991],

Conversely, a lot of intellectual work has to be done without the help of preexisting models and, as a consequence, do not benefit of efficient computer assistance. Barristers, for instance, need often to build up possibly different interpretations of document materials, to study their interdependences, to compare them, gather and classify information, produce specific documents from existing ones. Drafting tools and ideas managers are sometimes used to help such activities but they are independent tools more dedicated to document preparation than to knowledge elicitation. Hypertext systems, especially those integrating an interactive authoring environment, can assist such tasks in a more complete manner, bridging the gap between source documents and the production of new information because of their ability to incrementally handle knowledge elicitation.

Page 48

The purpoce of his paper is not to describe or propose yet another hypertext tool for the law domain1 but to examine at a general level why and how hypertext tools can assist the kind of intellectual work identified as information gardening which radically differs from information miming and manufacturing [Bernstein 1993]. We show that the classical hypertext paradigm must be enlarged and replaced by a more functional view in order to support this kind of activity. The different metaphors of mining, manufacturing and gardening/farming are situated in this new framework. We then examine how gardening/farming is supported by current systems. Finally, we show, as illustrated by the Mac Web system we have developed, how hypertext tools should contribute to make interpretation structure emerge from documents thus providing means to exploit information from these documents.

@2. The Hypertext paradigm Revisited

@@2.1. What is the essence of hypertexts?

Beyond the initial idea from Bush [Bush 1945], the hypertext notion has largely emerged from very different hypertext, applications or systems. Thus, whereas a lot of computer applications as different as an interactive encyclopedia on CD-ROM or a writing environment are considered as hypertext applications, a unanimously accepted lorrnai definition of the notion of hypertext is not yet really available. Hypertext is a paradigm.

From a naive and external point of view, hypertext systems are considered as document displayers thai offer a point and dick user-interfaces for-navigating in a document space. Browsing rrom one document to another has become a commonplace. But this is not sufficient to characterize hypertexts.

From an abstract point of view, the link-node paradigm is generally proposed as a basis for characterizing hypertexts. However, links and nodes are complex notions that may correspond to very different entities. Elements as different as Hypercard cards, Guide expandable documents [Brown 1989] KSM pages [Akscyn 1988], composites [Gronbaek 1994], Sepia activity spaces [Streitz, 1992], MacWeb virtual documents [Nanard 1993] are kinds of nodes.

Page 49

Similarly, hard-wired links, computable links, queries, slots in frames are considered as links. Other approaches based on sets suggest alternative to linking [Parunak 1991]. Furthermore, an evolution towards sophisticated models is observed. Using types on nodes and links and even frame-based approaches [Marshall 1992, op. cit.] are more and more frequently used to handle structure. This complexification makes the hypertext notion difficult to formally characterize.

Tuned in 1990 and re-edited in 1994 [Halasz 1994], the DEXTER2 hypertext reference model is now the most accepted model. It results from a cooperative effort to unify and generalize the work done by pioneers developers of hypertext systems. It is an attempt to capture both formally and informally the important abstractions found in a wide range of existing and future hypertext systems. Beyond clarifying the distinction between the storage and the run-time aspect of hypermedia systems as well as the connection with external applications handling documents, DEXTER supersedes the traditional node/link model by the notion of components. It emphasizes an implementation independence allowing various data models.

However DEXTER only characterizes the structural aspects of hypertexts and their consequences on hypertext architecture. Whereas it suggests a good architecture for robust implementation as illustrated in the Devise system [Gronbaek 1994, op. cit.] DEXTER only aims at observing hypertexts from the inside. It describes how hypertexts internally work but not the function they play in terms of human-computer interaction. Hypertext are described from a static point of view. The dynamic aspects of structure evolution during a knowledge acquisition process are not the concern of this model.

Thereby, an important question is "what is the role of linking or more generally of stablishing relationships between pieces of information presented as documents? What can we do with it?". Hypertexts need to be observed from a functional point of view.

@@2.2. Metaphors shaping information management in hypertext systems

Whereas hypertext systems technically rely on handling or using relationships between information pieces, they deeply differ depending on the kind of use of relationships and the style of application which is addressed.

Page 50

Three main categories of activities can be observed when a human is dealing with information:

- Extracting information from large repositories,

- Organizing existing information for presenting and better accessing it (and even selling it),

- Making information and structure emerge.

This diversity is stressed by Mark Bernstein [Bernstein 1993, op. cit.] who proposed a set of three relevant metaphors: information minings information manufacturing and information farming. These metaphors account for the attitudes and the interests of users when concerned with information.

They make clear what kinds of features are suitable in hypertext with respect to the addressed applications.

@@@2.2.1. Information mining

Information miners are interested in providing relevant information pieces found within large repositories more than in adding value to information or in creating it. Information is like nuggets extracted for subsequent use. Such approaches are mostly concerned by the cost-efficiency of extraction. They are very close to information retrieval. So, recall and precision are important criteria. Hypertext features improves the efficiency of retrieval: links are used for instance to record explicit relationships asserted by the user on retrieved data, in order to improve the efficiency of later searches [Croft 1989], Information mining is a frequent need in legal applications where, for instance, case law leads to retrieve documents about similar situations.

@@@2.2.2. Information manufacturing

Contrary to information mining which is more concerned with the relevance of the extracted nuggets, information manufacturing mainly focuses on ease of access to information by end-users. Information manufacturers refine raw information into high added value products specifically adapted to some end-users need. Hypertexts are well suited tools for presenting refined information. Relationships mapped onto information make closer information pieces and draw highways for accessing them more efficiently in context.

Legal tests are highly complex and benefit from information manufacturing. Relevant information is often disseminated in many documents. Building a rich structure for efficiently accessing them obviously providesPage 51 an obvious commercial added value. Hypertext on for instance fiscal texts, laws or decrees are very helpful and boost the efficiency of their users. Relationships between items are made explicit and semantically connect pieces of information, thus allow a reader to get all relevant information in a given context by a few interactions.

All the efforts in information manufacturing consist in designing and building consistent structures dedicated to the inteaded use of information. In addition to techniques that are suitable to take into account the specificity of non-linear reading, notions such as schema and some design techniques which have proved their efficiency for databases are helpfully adapted to hypertext [Garzotto 1993] and serve to producing large hypertexts, Like in industrial manufacturing, standards are important to preserve the investment for the long term, SGML [ISO 1986] and Hytime [ISO 1991] make it possible to keep the structure (that is mapped onto documents and provides the added value) in a form which is independent of any system and preserve it for the future.

@@@2.2.3. Information gardening/farming

The third metaphor. Information gardening, seems less obvious but is also...

Per continuare a leggere

RICHIEDI UNA PROVA

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT