Intelligent information retrieval from multiple databases

AutoreJaap C. Hage/Maarten Van Der Meulen/Georges Span
CaricaUniversity of Limburg, P.O. Box 616-6200 MD Maastricht (Olanda)
Pagine149-164

Jaap C. Hage/Maarten Van Der Meulen/Georges Span University of Limburg, P.O. Box 616-6200 MD Maastricht (Olanda) The research for this paper was partly financed by the Foundation for Knowledge Based Systems (SKBS), which seeks to improve the level of expertise in the Netherlands in the field of knowledge based systems, and to promote the transfer of knowledge in this field between universities and business companies.

Page 149

@1. Introduction

Modem information technology provides human workers with enormous amounts of information. However, this information is not always tailored to the needs of those working with it. Moreover, it is often necessary to consult more than one database to obtain the needed information. To combine the information from multiple sources, the human user must perform (mental) cut and paste operations [Greenes 1991]. But before arriving at the cut and paste-stage, the user must first be able to find the pieces of information which belong together, If the information is to be obtained from multiple databases, this may cost considerable effort. In short, retrieval of related information from multiple databases can be a troublesome affair.

To diminish these troubles we can use an environment that enables its users to have uniform access to multiple databases, that facilitates finding pieces of information which are in some manner mutually related, and that adapts its retrieval facilities to the needs of the user. The purpose of the Archimedes project is to offer an environment that supports uniform access to multiple databases, and that provides user-customized facilities to retrieve multimedia information.

This environment Is the so-called information network, which contains linked information objects pointing to pieces of information stored in multiple databases. The network offers graphical browsing facilities, next to the possibility to formulate queries based upon both types and attributes, of information objects and on the semantics of the links connecting them.

Page 150

In connection with the Archimedes project, a set of four applications will be built, which exploit the facilities of the information network. To demonstrate the capabilites of the Archimedes approach, the applications are taken from different domains, to wit a medical, a legal and an economical domain. We will illustrate the uses of the network by taking examples from these domains.

The Archimedes project is a project in progress. This paper describes the design of the information network which is the central topic of the project. In the next section we start the description of the Archimedes approach with a discussion of the information objects. In section 3 we describe how the information objects are connected in the information network, The fourth section is about information retrieval from the network. Then follows an account of the role and nature of user models. And finally, before concluding the paper, we indicate in section 6 how it is possible to include the knowledge of a knowledge system in the information network, so that the knowledge can be integrated with the other information.

@2. Information objects

@@2.1. The purpose of having information objects

One starting point of the Archimedes approach is that uniform retrieval of information from multiple databases should not go at the cost of the original databases. The data should remain in its original form. Since the original form may be different for different databases, and since we want to profit from the advantages of the object oriented paradigm [Parsaye et at 1989; Roberts 1991], we decided to work with object-like counterparts of the original pieces of information [Akscyn et al 1988; Noll and Scacchi 1991], These counterparts are what we cai information objects.

In this discussion of the information network and information objects, we approach the network from the user's point of view. The object oriented paradigm can not only be applied to the structure of software, but also to the picture the user is given of an application. (This picture is confusingly also called the 'user model", just as the system's model of the user), When information objects are discussed, we refer to objects in the eye or ear of the system's user, Implementation details are not discussed in this paper.

In using information objects, we hope to achieve the four types of transparency a heterogeneous information environment should provide according to [Noll and Scacchi 1991]. There should be type transparency,Page 151 which means that the user should be able to manipulate an object independent of its type. There should be location and source transparency; the user need not be aware of either the distance or the type of the information repository. And finally there should be scale transparency; objects should behave the same regardless their number,

@@2.2. Types of information objects

Information objects come in three major types. First there are the counterparts of pieces of information which reside in the original databases. We will call these counterparts data objects. Since the original databases are not contained in the information network, they are called external databases. Data objects may refer to data of multiple kinds, such as text, graphics, sound, and animation.

The data objects do not contain the information themselves. Next to the attributes and methods they need to fulfill their role in the network, they have methods to retrieve the data from their original sources and display it. The latter methods vary according to the nature of the external databases. A relational database, for instance, is accessed differently than a database containing unstructured texts, graphics or sound. The methods employed to access the external databases take care of the necessary transformations and may therefore be compared to mediators in the sense of [Wiederhold 1992] (fig. 1).

Figure 1. The role of mediators

[ GRAPHICS ARE NOT INCLUDED ]

Data objects are typed, so that the type reflects the nature of the data the objects point to [cf. Diaper and Rada 1991]. The types of the objects can be used in queries.

Page 152

Besides data objects there is a number of related concepts. The concepts contain no information, but through their mutual links and their links to other information objects, they facilitate information retrieval from the network and make conceptual retrieval possible. The conceptual part of the information network may be seen as the cement of the network which binds the objects together. We discuss this role of the concepts more extensively in section 3.

Finally the information network contains elements of knowledge. The information-network as a whole contains knowledge in two ways. The first way is based upon the types of the information objects and the links between them, which make it possible to draw inferences about the relations between information objects [cf. Diaper and Rada 1991]. The second way is based upon information objects which contain knowledge themselves, such as, for instance, rules. In section 6 we discuss the role of knowledge objects more extensively.

The relationship between the three types of information objects on the one hand, and the external databases on the other hand, is pictured in fig. 2.

Figure 2. The network in relation to external databases

[ GRAPHICS ARE NOT INCLUDED ]

Page 153

@@2.3. Compound and elementary objects

Information objects are elementary or...

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