BRAQUE: A Hypertext-based Interface for Accessing Large Text Databases

• Autore: Pier Giorgio Marchetti
• Pagine: 101-112

Page 101

@1. Introduction

The amount of literature indexed in bibliographic and text collections continues to grow, is complex and multidisciplinary in nature. The access to such sources of information, their consultation is required by more arid more professionals. Tools like personal computers and high speed networks to connect to on-line catalogues and bibliographic collections are easily available. Nevertheless a professional knowledge of collection contents, indexing methods and query language is normally required to allow a fruitful access to large text collections and retrieval of relevant information. This paper addresses the personal computer version (Ms Windows) of the BRAQUE (BRowse And QUEry) interface designed to overcome the above mentioned problems and give significant results in the information gathering process to the professional information hunter as well to the occasional searcher.

BRAQUE relies on the implementation of a two level hypertext model to support the hypertext metaphor in a very large document space supported by meta information (concept space). Even thought the underlying model is valid for any kind of information object (including compound documents containing graphs, pictures or sound), previous experimental results and current implementation refer to text information objects. The validity of the BR000E approach, therefore includes the general case of compound information objects (compound documents) but there is room for further development and exploitation of the model and of the related implementation.

@2. BRAQUE: Project Background

The BRAQUE interface has been designed trying to match two completely different sets of requirements. The first requirement was to satisfy the needs of as many user profiles as possible. in the past this requirement got solutions which were trying to categorise users according to their professio-Page 102nal skills, and then to tailor interface behaviour according to such categories like ´expertª, ´noviceª, which proved very rigid and unproductive. Following a cognitive approach [Ingwersen 1984] a number of efforts has been made up to an explicit use of Cognitive Task Analysis (cta) [Roth Woods 1989], with the attempt to identify the actions performed by individuals in information access.

The initial design effort [Belkin Marchetti 1990], used the Cognitive, Task Analysis of information retrieval (ir) interaction, based on the distributed expert problem treatment model developed by [Belkin et al. 1983].

In the earlier work the analysis had to rely on the specific situation of user-intermediary interaction in document retrieval.

Some insight was gained, but after some analysis it became clear that user behaviour was influenced and changed according to the information problem or according to feedback from the interaction with the (digital/on-line) information sources. Several attempts to perform synthesis and design starting from user interactions and CTA failed mainly because users of information systems try to satisfy their information requirements with several information seeking strategies. It is therefore extremely difficult to design an user interface pretending to use the insight from CTA or from user interaction records and try to generate a corresponding finite state machine. As a matter of fact, the (users') information seeking strategies are influenced by the nature of the information problem, by the nature of the information itself (e.g. information, meta-information), by the user's goal (e.g. learn, select), by the information access method (e.g. scan, search) or by the information access mode (e.g. recognise, specify) [Belkin Marchetti Cool 1993].

So the design specification of BRAQUE was tailored around the aim of providing users with at least some of the information seeking strategies mentioned above at any time.

The second requirement was to exploit the concepts categorisation performed during the document indexing process, to provide feedback both in the concept and document browsing, and in the querying of the document collection. in this view the information retrieval process can be represented according to the schema in Fig. 1 which extends (including, in an explicit way, concepts indexing, thesaurus and feedback) the information retrieval schema according to [Belkin Croft 1987].

A conceptual modelling of IR data resources, through a two-level description of IR systems, had been used to describe and model the functionality of an hypertext environment for interacting with large textual databases [Agosti et al. 1991, 1992]. The hyperline prototype of this functionalPage 103 schema has been under test for a long time in an operational environment [Marchetti Muehlhauser 1991], The list and a schema of the atomic functions which...