Computer systems analysts/designers have for years used diagrammatic "languages" or grammars to represent, model, and confirm users' requirements for new information systems. Since the 1950's, diagrammatic tools such as systems flowcharts and program flowcharts have been used by analysts/designers to model computerised solutions to problems in the real world. Moreover, the diagrammatic techniques predominantly used have changed many times. The world of systems analysis and design has moved from using systems flowcharts and program flowcharts through to the use of data flow diagrams, entity-relationship models, and structure charts to embracing now object-oriented analysis and design schemas.
Each major change in this transition has been motivated primarily by people needing to overcome real-world representation deficiencies in the previous generally-accepted technique(s).
Moreover, each major change has associated with it significant costs for business, viz., the training and re-training of personnel in the new techniques, the increased documentation requirements of the new techniques, and the re-engineering of older systems under the requirements of the new techniques. It is important then to business to minimise the need for these transitions. Accordingly, do object-oriented schemas represent the ultimate conclusion to this modelling transition path?
It is important to have a theory that explains the set of real-world constructs a grammar must be able to describe. In the absence of such a theory base, Olle et al. (1991) detail the proliferation of symbols and grammars that could be used for information systems modelling. By contrast, armed with such a theoretical foundation, analysts/designers could analyse and evaluate modelling grammars as they become available. Strengths and
weaknesses of the grammars could be determined. Finally, over time, a complete modelling grammar may emanate. Only then could this cycle of moving from one set of modelling grammar(s) to the next, and the significant costs associated with such transitions, end.
Since the early 1980's, automated support for modelling in the form of Computer-Aided Software Engineering (CASE) tools has been available to analysts/designers. The manufacturers of such CASE tools have implemented many grammars available. Furthermore, in a number of instances, they have added to the number of symbols already provided within those grammars. Accordingly, in such environments, a theoretical foundation that established the set of real-world constructs needed could assist analysts/designers in selecting the grammars and symbols they need to perform their modelling tasks. Moreover, such a theoretical base could assist analysts/designers in evaluating and selecting a CASE tool that most
appropriately supported their modelling needs. Finally, CASE manufacturers could use such a theory to rationalise the number of grammars and symbols they are called on to provide and thus, minimise their production costs particularly in terms of consistency-checking utilities.
This work has proposed and used Wand and Weber's (1989a, 1989b, 1990a, 1990b, 1991, 1993, 1995) models as the theoretical base required for the specification and evaluation of information systems analysis and design grammars as implemented in modern CASE tools. These authors have spent much time defining and articulating fundamental information systems concepts. Accordingly, their models were used to analyse the grammars provided within a structured CASE tool, EXCELERATOR VI.9. From this analysis, hypotheses regarding the use of grammars within the CASE tool are generated. Because there may be other factors impinging on the grammar-use decision, a model of decision making within a
software-aid environment (Nevell 1991) was adapted for this research. Other minor hypotheses were generated from this model.
The hypotheses were tested using a multiple-method approach involving a survey of EXCELERATOR users (168 responses) and structured interviews (36 interviews) with selected users. Ontological incompleteness in grammars was a strong influence on the decision to use combinations of grammars when modelling in the CASE tool. Other strong influences on the grammar-use decision were a perception of minimal problems with maintenance of models and a perception of minimal problems with consistency of representation across models. The concept of minimal ontological overlap (MOO) was operationalised in this work in an attempt to predict the combinations of grammars that analysts/designers might find most useful. While the quantitative results showed that MOO was not dominant in predicting the combinations used, qualitative
insights gave encouragement on the usefulness of MOO for establishing grammar use by analysts/designers. Irrespective of the grammar combination used, however, analysts/designers did not appear to change significantly their preferred grammar set when organisational standards in this area were removed. Finally, perceptions of minimal problems with consistency of representation across multiple-grammar models are significantly influenced by the recognition of the existence of an automated integration mechanism within the CASE environment. While not a significant influence, there was encouragement for the view that the use of an MOO-conforming grammar set might also lead to perceptions of minimal problems with consistency of representation.
In summary, this research has contributed to the field of information systems analysis and design by giving insights into why and how analysts/designers use grammars within CASE tools. This work has identified a promising
theory base for specifying and evaluating ISAD grammars - the BWW models. It has clarified and extended a number of the constructs and concepts within the theory. It has applied the theory to the use of grammars in combination. Furthermore, this theoretical analysis of the use of combinations of grammars has been extended into the CASE tool environment. Finally, this work has provided a substantive empirical investigation into the usefulness of the theoretical foundation.