My Ph.D research (1994-1997) was supported by a Ph.D. studentship from the Greek State Scholaship Foundation
An Agent-based Hypermedia Digital Library
Michail Salampasis
Abstract
Nowadays, the number of computer networks proliferates and a vast amount of diverse information sources are becoming increasingly available. The term Hypermedia Digital Libraries (HDLs) may be used to describe these highly dynamic, interactive and distributed information seeking environments. In HDLs information seeking is possible using combinations of opportunistic browsing and analytical, query-based strategies. The advent of HDLs posed several information seeking and architectural problems which must be properly considered before a HDL can be realised efficiently and effectively.
One problem that information seekers face in any distributed electronic environment is the collection fusion problem. This term is used to delineate the problem of selecting information sources from the many available and, the production of a single merged result which can be effectively examined. Several techniques have been proposed to solve the collection fusion problem. These techniques can not be utilised in dynamic and large environments such as HDLs easily, because they require a learning phase or excessive exchange of information. From an architectural perspective, another problem is the difficulty of conventional, "closed" hypermedia architectures to support the design and development of HDLs. Even distributed hypermedia systems like the World Wide Web have several limitations commonly found in "closed" hypermedia systems.
In this Ph.D. thesis a new collection fusion strategy is presented and systematically evaluated which facilitates hypermedia links to solve the collection fusion problem. The link-based strategy is applicable in dynamic HDLs because it does not require any learning phase nor uses excessive information to solve the collection fusion problem. Also, a novel distributed agent-based Open Hypermedia System (OHS) is presented and evaluated. The agent-based OHS can be used as an underlying platform for designing and developing open, interoperable and extensible HDLs in which multiple information seeking strategies can be integrated.
