In this thesis, we examine algorithms for content selection and information presentation that enable browsing as an information-seeking strategy. Our hypothesis is that users who wish to explore the contents of a particular domain might not be satisfied with traditional spoken dialogue systems, in which focal information is not described until a small number of the records comprising the domain are identified through a constraint-elicitation process. Specifically, we hypothesized that users would prefer responses in the form of incremental summaries of the focal information.

Database Management Systems (DBMSs) support the creation and modification of databases containing potentially very large collections of data. Typically, a DBMS embodies or enforces a data model. The most widely adopted data model remains the relational data model proposed over three decades ago by E. F. Codd, and it is envisaged that the relational model will remain the most popular model in the future.

This thesis describes the formulation and experimental use of psychological principles that apply to conceptual modelling as practised during information systems development. The principles address cognition (perception, memory and mental models) and group dynamics. The aim is to determine whether application of fundamental psychological principles can help to make modellers, especially those who are relatively inexperienced, more effective. An experimental graphical modelling technique (method 'X') is presented that conforms to the psychological principles, together with a supporting software tool for visual construction of models in the design of typical business database systems. The effectiveness of both inexperienced and expert modellers using method 'X' in real business situations was compared with that of modellers using conventional object modelling. Data was gathered in a series of field experiments using participant observation, questionnaires, and interviews and by analysing the resulting models. With conventional object modelling, untrained modellers produced results that were grossly incomplete and incorrect (22-35%, on average). Using method 'X', untrained modellers produced models that were almost complete and correct (better than 82%). Significant productivity gains were observed with method 'X' (approximately 150% for expert modeller and over 450% for untrained modellers). For an expert modeller no measurable differences in quality were observed between methods, but the modeller regarded the quality of method 'X' models as better and expressed a preference method 'X' over the conventional approach. The results appear to support the idea of re-engineering conceptual modelling practice according to psychological first principles. The fact that more dramatic performance improvements were observed for inexperienced modellers suggests that modelling need not require a high degree of expertise, if methods and tools are adapted appropriately. The results could be exploited to empower untrained modellers, such as end users, who wish to develop large software systems but lack access to the skills of trained IT professionals.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

No description available.