Task-oriented access to data files: An evaluation

Abstract

DalText, a prototype information retrieval system has been developed that permits the user to select access methods based on the task at hand. The system integrates multiple access techniques to such textual data as those generated through word processing packages, newsgroups, e-mail, etc., that are not maintained in traditional database management systems or information retrieval systems. The access methods available include viewing the data as a sequence of records for browsing, generating sets of records through string matching and Boolean combinations, and through generating a table schema and instantiating attribute values dynamically. As the access methods are all based on the same underlying data access model, the user can flip back and forth between the access methods in order to best accomplish the task at hand. An evaluation was conducted to test for correlations among a number of different variables. Of note among the results, it was found that the subjects used the access method best suited to the information task rather than relying on the access method with which they were familiar. The results indicate that further investigation should be conducted on the development of information retrieval systems that allow users to select access and display methodologies appropriate to the task at hand. IntroductionTraditionally, information systems have been developed to provide effective and efficient use of given data sets. The models used for the development of these systems have been based on characteristics of the data rather than on characteristics of the information needs of the users (Kuhlthau, 1991). For example, information retrieval (IR) Received June 1, 1993; revised November 2, 1993; accepted November 2, 1993. This research has been supported in part by the Natural Sciences and Engineering Research Council of Canada Operating Grant OGPOO09163 and by the Research and Development Fund of Dalhousie University.*To whom all correspondence should be addressed. systems (Salton, 1989) have been developed to provide users with access to relatively unstructured textual data based on the properties of word occurrences, while database management systems (DMBS) (Date, 1990) have been developed to provide users with access to more structured data, such as employee or corporate data, based on properties and values of well-defined attributes or fields within certain domains. Typically, users map their information needs into the query language of the specific system and then interpret the results within the framework of their needs. This is the data-centered paradigm for information access (Watters & Shepherd, 1992) and is based on the premises that information needs can be expressed or reformulated as distinct queries and that all needs from a data set can be satisfied by one form of output (i.e., records of text or tables of values).This approach is limited, however, as it does not take into account either the user or the task the user is attempting to perform. At the beginning of a search session, users oft...