BUR99373
Applying Theory in Practice: Assessing the Information Requirements of an Organisation. ®
Oliver K. Burmeister & Peter Eden
Abstract
Data is capital & has value to an organisation. As such it needs to be adaptable & useful over time to that organisation. Data that is designed to reflect users' understanding is more useful and adaptable to changing conditions. Yet how can such significant concepts be brought across to students learning the theory of database design? Once the data requirements are collected and analysed, there is a well understood theory that can be applied to database design. This is the focus of most first courses on database. The theory can be, and often is taught in abstraction, removed from the practicalities of how the data is understood by users in real industrial settings. Not so at Swinburne University of Technology where there is an ethos of employment focus in all teaching, from TAFE through to post-graduate education. The pedagogy is informed by industry & in large part based around case studies of practical problems. With particular emphasis on conceptual modelling, students analyse real world situations and design database solutions. Students don't just learn the design theory but the whole process of problem analysis, classification and solution. They employ valuable analytic skills such as abstraction, pattern recognition and validation to develop their database design. This approach empowers students to develop skills and understanding beyond the narrow confines of idealised database design, and in the process, develop an appreciation for the typical business problems that software technology can help solve. The case studies make significant use of sample data to help validate the analysis. A feature of the approach is to guide the student through a series of steps that require a rationale at each stage to justify the modelling decisions. Such a rationale forces the student to consider the needs of the business as well as the structure of information.
Communication is a social process that takes place in an environment of shared information. Database technology is specifically directed at providing computerised support to manage bodies of information shared among people. Databases have been particularly significant in supporting the need of organisations and businesses to hold large amounts of highly structured information. The integrity and availability of such information in computerised databases is now critical to the functioning of many human organisations.
The successful application of database technology depends on an accurate assessment of information needs. To that end, the organisation has to be examined from the perspective of information. A thorough description of the information held by the organisation must be available it the technology is to be usefully deployed in the long term.
The now commonplace availability of computerised database management systems (DBMS) across the range of computing machines coupled with the recognition that the value and lifetime of stored information soon exceeds the systems that manage it, means that database design issues have a broad impact on the user community. Fortunately these problems are well understood and their solutions make for interesting and relevant core material in any undergraduate curriculum with a computing minor.
Communicating to a general audience the issues of information in the organisation as well as the technical substance of database design, a problem solving approach based on concrete situations is absolutely necessary. This paper describes and discusses an approach taken at Swinburne University of Technology (Swinburne) to a first course on Database founded on analysing information requirements by method called conceptual data modelling. The method is not new, but its inclusion in undergraduate curriculum is unconventional.
The paper begins in section 2 by discussing the organisational facilities and constraints of the Database subject, at the University, School and Course levels. In section 3, the elements of database analysis and design are described, including a critique of what is used in typical undergraduate courses. In section 4, there is a review of the various approaches described in the text book literature to using semantic data models. As a response to the issues already raised in the preceding sections, the distinguishing features of the data modelling method used at Swinburne are described in section 5. In section 6, the educational response to the delivery of the content is examined. Finally in the last section, we conclude.
The focus on conceptual modelling in the first course in database teaching at Swinburne University of Technology is a direct result of the ongoing industry practice strategic orientation of the university. Swinburne has a long history of vocational orientation for students that either directly involves students in cooperative educational opportunities with employers, or through employers on course advisory boards, which focuses studies toward employability. This was evidenced in a 1994 Federal government report of employment of university graduates (Pristel, 1994; Farouque, 1994). 67.1% of Swinburne graduates found employment within six months of graduation - the second highest in the nation; the national average was 54.3%. Similarly, a 1995 survey of Victorian universities reported that "Swinburne University of Technology has the highest portion of graduates across a range of faculties who found a job within five months of completing their studies." (Richards, 1995, p. 6) This distinctive feature of study at Swinburne is what attracts many of our local students & also our overseas full-fee paying students. Market research time & again shows this to be the case. Hence the continuing emphasis in the current marketing plan for the University which says: 'A feature of many Swinburne undergraduate courses is the applied emphasis and professional orientation' (McAdam, 1999, p. 1).
In the late 1980s a senior appointment within the now School of Information of Technology was that of an industry practitioner of more than 25 years experience in IT. Her view was that the traditional foci of database courses did not adequately prepare students for the real world. It was her influence that first brought about the changes at Swinburne from the traditional database approach to that of conceptual modelling & in particular "Fact Modeling" (James & Olsen, 1994). Later, through the ongoing involvement of industry practitioners on various course consultative bodies, this emphasis has remained; though Fact Modeling is now reserved for Masters level study, whilst the first database course now has an emphasis on conceptual modelling through Entity Relationship Analysis (ERA) & resolving complex ERA through synthesis, i.e., through Functional Dependency Analysis. Even now, in positioning itself for the future, one of the university's strategic foci is that of industry partnership. For example the marketing plan for this year has in its opening sentence: 'The vision of Swinburne University of Technology Higher Education Division ... is to become the 'University of First Choice' for students who want a complete professional education' (McAdam, 1999, p. 1). This has long been a distinctive feature of a Swinburne education & is something that continues to inform the pedagogy of many of its courses.
At Swinburne Database 1 is a subject that appears in a variety of courses. The main contingent of students traditionally in this subject has been Business students, which has further heightened the need to emphasise the importance of the setting of many database applications in business. In the last 18 months as a result of economic course rationalisations across the University, this subject now also has become the first, and in many cases only, database subject for students from as diverse a course range as Arts, Humanities, Engineering & Applied Science. Though the majority of students are undergraduates, there are also large numbers of Graduate Certificate, and Graduate Diploma students; There are even Master of Engineering students who choose this subject as an elective. Within this diversity business students still comprise a significant proportion (approximately 35%) of the more than 1,000 students who take this subject annually. All these students have 4 contact hours of on-campus classes, consisting of a two hour lecture, one hour tutorial and one hour laboratory session initially, that part way through the semester changes to no laboratory session and two hours of tutorial work instead. This diversity in the student population has meant a number of changes in the curriculum so that each stream is adequately catered for. Yet one central theme remains & is the primary focus of teaching in the subject, namely "Conceptual Modelling".
Conceptual modelling forms a significant portion of the 12 week semester. Entity Relationship Analysis, which in other first databases courses often only has a single week of coverage here consumes 25% of the curriculum. Another 2 weeks are devoted to Functional Dependency Diagrams, another conceptual method. Even Normalisation by Decomposition, which consumes a further 2 weeks, is not taught solely for its own sake as in other first database courses, but also with a view to checking that the end result of conceptual models converted to relational schema are indeed in at least Third Normal Form. Thus effectively 5 out of 12 weeks directly concentrate on conceptual modelling and 2 further weeks do indirectly. This is an unconventional approach that emerges from the industry focus described above.
The subject emphasis is thus on gaining an understanding of a problem in its context and on the management of the organisation's data to ensure that the information produced by the database system is relevant and accurate. Students are taught to use conceptual data analysis methods to produce a logical data model and then to test their models by implementing them using SQL. This learning process is cumulative & summarise for students in a learning map (Yates & Chandler, 1994; Burmeister, 1995) that is referred to in every topic in the subject to facilitate their understanding of the cumulative subject matter & their integration of it into a holistic understanding of conceptual modelling. That learning map is available to students both on the cover of their Subject Learning Guide (Jeffery, Smith & Weal, 1998) & in their online education subject matter.
Anecdotally, a number of students returning from IBL have testified to the direct relevance of Database 1 to the work they have done. Two of these students volunteered to have their testimonials recorded & digitised for an online introduction component of the subject. The reason such was sought was to encourage students engaging in the subject to view right from the start the significance from an industry standpoint of the study they would be engaged in.
In this section, the context of conceptual modelling on database development is described.
Conceptual Modelling is an approach to software development based on the notion that software is fundamentally a human communication medium (Winograd & Flores, 1986) The behaviour of a software system is meaningful only by virtue of the interpretation made of it by users of the system, so that software ideally is constructed to meet the requirements and thinking of users. The justification of this approach is founded not only on getting the software to correctly reflect the users' requirements, but also on a deeper principle that a system that is engineered to reflect the users' situation in the business is more understandable and better able to accommodate change.
Conceptual modelling aims to create an abstract representation of the situation under investigation, or more precisely, the way users think about it. Indeed, analysing the domain of user concepts should be done quite independently of any software technology considerations. This is embodied in the Conceptualisation Principle (International Standards Organization, 1987). A conceptual model is initially a representation of the business situation using concepts meaningful only in it, not of the software to manage the situation.
So in both the analysis and design of software, the pattern of users' communication and the structure of information should be a focus of attention. Analysis is concerned with discovery and specification of the information system whereas design is concerned with its logical construction (Loucopoulos & Theodoulidis, 1992).
The aim of analysis is to prepare a specification that is a valid representation of the user's requirements as well as a consistent statement of the contract between the user and the developer. So conceptual modelling is accurately defines the extent of the analysis phase.
In modelling the requirements for a data intensive system, we focus on what information is necessary for the business to function. Information is not arbitrary, but describes things that matter to users in the business situation. As the real world situation is not in an arbitrary state, but constrained by natural laws and business rules, only certain combinations of facts can describe a possible world. Documenting, analysing and validating such constraints is an essential part of conceptual modelling for a data intensive system.
Data modelling requires a language and notation to specify and constrain data. Such a language is called a data model. A semantic data model (Borgida, 1991) allows the analyst to directly represent a world conceived as composed of objects and properties (Wand & Weber, 1989). A semantic data model has the capability of representing the structure of information about things in the real world, so is an essential tool in conceptual data modelling.
The Entity-Relationship data model is the most popular of the semantic data models (ACM/IEEE-CS Joint Curriculum Task Force, 1991; Australian Committee for Training Curriculum, 1994; International Standards Organization, 1991) so a brief account of its capabilities is described here. Classes of objects in the business are identified and modelled as entities. Entities have properties that are either attributes that directly describe entities or can be relationships that model facts about entities. Entities, relationships and attributes are depicted visually in an Entity-Relationship diagram. There is also a visual constraint language to model business rules.
The product of conceptual modelling in data analysis is a conceptual schema, or specifically an Entity-Relationship diagram.
The input to the design process is a clear and unambiguous statement of data requirements. In design, we are concerned with how the software will be organised to fulfil these requirements.
In starting with a conceptual schema, the designer is free to choose any way to fulfil the requirements because the schema is a necessary and sufficient specification of user needs. If the design of the software is based on a conceptual schema, the resulting system architecture should reflect its purpose in a way that makes it easy to adapt to changing requirements.
Most DBMS separate the logical view of data from the physical storage of data (Tsichritzis & Klug, 1978) by providing a language for defining data structure by a Data Definition Language (DDL) and manipulating data by a Data Manipulation Language (DML). A database design then amounts to a set of data definitions in the DML.
So, once a conceptual schema is at hand, the design process is governed by relatively straightforward procedures for implementing, say, an Entity-Relationship diagram in the appropriate DDL (Elmasri & Navathe, 1989).
Much attention has been paid to solving the problems of maintaining the integrity and accessibility of quantities of stored data. The result has been the emergence of database technology and relational database design theory. A design resulting from a conceptual schema can be demonstrated in the theory to have desirable properties such as minimal redundancy, being easy to update, query and extend.
The most complete theory of database design is founded on the Relational Data Model, (Codd 1971). Relational database design theory is predicated on eliminating certain so-called data anomalies that complicate the processing of data. Traditionally this is achieved by the process of normalisation. Normalisation delivers a set of storage structures that minimise the effect of data anomalies. At the same time, normalisation produces a representation of the business situation that is natural and understandable. So in this case, good data design largely coincides with a faithful representation of users' concepts about the business.
The input to the normalisation process are the so-called data dependencies. Much of relational database design theory is about the formal manipulation of data dependencies. The presence of data dependencies directs the normalisation process. Relational theory assumes data is stored in relations or tables. In its original form, normalisation is a design strategy that incrementally improves an existing database design through various levels, known as first, second, third, fourth etc. normal forms (Kent 1983). This process is called Normalisation by Decomposition because in progressing from one normal form to the next, larger tables are decomposed into smaller tables.
Normalisation by decomposition is appropriate once some storage tables have been proposed and the design needs improvement. This method is predicated on an assumption that a universal relation exists from which all other tables are decomposed (Atzeni & Parker 1982). The universal relation does not always exist in practical situations and if so, it is not usually a natural construct. Normalisation by decomposition also suffers from the problem of nonunique decomposition (Dutka & Howard 1989). Beyond Third Normal Form the resulting design can depend on the order of decomposition. This violates the Conceptualisation Principle because the designer is forced to make a choice of design while still analysing the problem. The relational data model is not a semantic data model, so is not sufficiently rich in constraint types to specify the data structure without requiring design choices.
Traditional database design curriculum is firmly based on relational data design theory. Conceptual data modelling is usually considered to be an advanced topic, so as such is not usually treated in the undergraduate curriculum. An examination of typical undergraduate texts on database (Elmasri & Navathe 1989, McFadden & Hoffer 1994) shows they focus largely on the logical level and employ normalisation by decomposition as the primary design technique. Very little attention is given to the heuristics of problem solving required to convert an unstructured requirement into a structured specification. The Entity-Relationship data model is used only to depict a design either already conceived or as a very high-level, coarse view of the data requirements. It is not used as a vehicle for rigorously recording and analysing data requirements directly from the problem domain.
There are perhaps a number of reasons that conceptual data modelling is not significant in the tradition undergraduate curriculum. First, logical data models such as the Relational, Network and Hierarchical models in the early 1970s historically predated the Entity-Relationship data model (Chen, 1976). The theory of semantic data models was not presented until the late 1980s (Peckham & Maryanski,1988) and very few were implemented. Relational DBMSs became very well accepted in industry and in education through the acceptance of the SQL standard and the growth of microcomputer systems. Consequently database software and how to program the database became the focus of attention. The attraction of students implementing databases at the logical level in software distracted from the important task understanding and analysing the business problem.
Another reason for the absence of conceptual modelling from the undergraduate curriculum is that computer science has traditionally been the foundation of computing courses. The emphasis was still on programming the utilities that the machine provided, rather than understanding and solving the problems of human interaction, communication and language (Winograd & Flores 1986).
The situation was somewhat better for relational databases because of the simplicity of the logical data model and the well developed theory of design, normalisation. The relational data model narrowed the 'semantic gap' between the problem and the solution domains and was adequate for small problems. It was assumed that those methods could just be scaled up to solve larger problems. So it was, but only by people who already had a lot of experience in database design. Students came away with a theory and a formalism that they could apply only to simple, somewhat artificial problems.
A notable exception to the above is that of NIAM (Halpin, 1995) for the Object Role data model which defines a comprehensive process for conceptual data analysis, commonly called fact modelling. Fact modelling has been introduced at some Australian universities (Edmond, 1992) at the undergraduate level in the past, but the notation is not widely understood, uses unconventional and voluminous terminology and really requires a whole semester to do it justice.
We identify three variants of the process of data modelling from the text book literature.
Data modelling is presented as a deterministic process where there is a single correct outcome. Often the procedure for arriving at the unique solution is not explicated. In practice the analyst tests various design alternatives, but the reality of such experimentation is rarely acknowledged. The proof of the solution is only in its implementation in software. It is common in the pedagogical database literature for an Entity-Relationship diagram to be used as a vehicle for explaining the data model constructs rather than how the diagram was derived in the first place. The semantics of the Entity-Relationship is presented as only loosely based on instances in the business.
Data modelling is presented as a creative design activity (Simsion 1994) where the element of choice is acknowledged, and the effect of decisions made by the modeller have design consequences. This approach tends to blur the distinction between analysis and design. Analysis and design progress iteratively. While it is sometimes the case that in order to unambiguously specify information requirements, it is necessary to make a design commitment, according to the Conceptualisation Principle, any concepts or constructs that go beyond what can be directly validated by the user should be avoided. Usually there are many ways of designing a requirement. Here is analyst is conscious of the design consequences of his specification.
Ideally any two analysts given the same interpretation of user requirements should produce identical specifications, however in practice, for various reasons, this may not happen. For instance, one specification may have made different design assumptions from the other. Or one specification may be notated differently, but equivalent to the other. This approach develops a specification from user requirements by a series of explicit transformations activated in specific contexts. Thus each step in the development has a rationale. Any different but equivalent specifications should be transformable from one to the other. A method that guides the student analyst through data analysis by a series of transformation is advocated here. This approach (Eden, 1996) combines the variety of the Iterative Design with the determinism of Identification and Documentation. Schema transformation has been used (Batini, Ceri & Navathe, 1992) as a developmental aid, but the semantics of such transformations are not defined, nor they organised into a coherent method. Schema transformations have been extensively used in fact modelling (Halpin, 1995).
For a full description of the method of conceptual modelling used in a first course of Database at Swinburne see Eden (1996, 1997). Here only the main features will be listed.
The method begins by an examination of a descriptive text and sample forms. By looking at parts of speech, candidates for entities, relationships and attributes are posited. Various tests for the validity of these candidates are applied and composed into an Entity-Relationship diagram. The diagram is labeled so it can be verbalized as sentences types. The importance of linguistic analysis was recognised by Abrial (1974) and Chen (1983).
The semantics of a semantic data model is defined by populations of facts, so an Entity-Relationship diagram is meaningful only in that it represents facts about the business. The method requires the analyst test that the diagram is meaningful and valid by populating it with facts. Every entity represents a population of instances and every relationship represents a population of facts about entity instances. This is an essential 'reality check' of the diagram. This has been advocated by Simsion (1994) and Halpin (1995).
A major source of difficulty for novice data modellers is to analyse information that depends on, or is about other information. For example, if a number of people attend a meeting, the people can be identified by their names, but the meeting itself is information that a room has been reserved at a particular time. Attendance cannot exist without a meeting or people and meeting needs a room and time. Entities with an independent existence are called strong, otherwise they depend on other entities and are weak.
Initially the method very strictly steers novices away from weak entities towards strong entities. This provides a developmental direction and is the backbone of the method. Depending on the context of the diagram, weak entities are constructed in a disciplined way by transformation, progressing from less detailed to a more detailed diagram.
The grading of entities by strength allows the novices work to be more easily diagnosed and assessed. The transformational approach provides an aetiology of entities. This is valuable for both the student and the instructor.
Apart from some exceptions (Date, 1986; Rock-Evans, 1989) entity strength is not widely employed.
Another difficulty for novices in data modelling is to precisely define what is one instance of an entity. There must be sufficient attributes of an entity to allow any two instances to be distinguished. This requires assigning keys to each entity, but not for the purposes of implementing a primary key, but to precisely define what things can populate the entity. This is easy for some kinds of entities such as people. Strong entities have a simple identification scheme and an independent existence. On the other hand, weak entities may have an identification scheme that depends on that of other entities. The method insists that every entity is identifiable. This imposes a level of rigour on the diagram that makes it easy to populate and thus validate.
Identifiability is recommended in only few texts (Yourdon, 1989; Barker 1990; Carter 1995; Halpin, 1995).
The process of expanding the Entity-Relationship diagram from strong to weaker entities is considerably more disciplined by the introduction of functional dependency diagrams to analyse the structure of closely related attributes. It is at this stage that the dependence of weak entities can sometimes be quite complex, so the Entity-relationship data model is not able to specify the constraints that direct the expansion. The use of functional dependency diagrams as a complement to Entity-Relationship diagrams is a major feature of the method described here.
Process-based learning refers in this instance to the pedagogy employed in Database 1. As can be seen from the discussion above, the traditional approach to first courses in database theory can involve a deterministic process where there is a single correct outcome. Such an approach can mean that students can achieve satisfactory results purely through the use of rote memorisation techniques. That is, the learning tasks do not demand that students appropriately structure and integrate their learning in the subject. Not that this is the intention of such courses, nor is it encouraged in such courses. However in the Database 1 subject rote learning would lead to failure, by the nature of the conceptual modelling tasks that students are set during assessment tasks.
Structuring strategies are a necessary part of learning in the subject. This flows naturally out of the problem orientation and is reinforced in the resources made available to both tutors and students. The focus of the learning is on the process of deriving appropriate solutions and on being able to argue in favour of one approach as opposed to another, as is often necessary in group projects undertaken during the semester, so that the whole group is agreed on the appropriate process for solving a problem.
Process-based learning is a pedagogy that is informed by a number of logistical and education factors. In the main these are problem-based learning, Multi-Modal Learning and learning structures.
Process-based learning is in part a pedagogy that involves problem-solving. As shown above, conceptual modelling is 'problem oriented' rather than 'solution oriented'. The aim is to record as faithfully as possible what the analyst observes about the business. For this reason process-based learning has its roots in problem-based learning, which in turn has its roots in professional education.
In the professional world problems tend not to be organised in a fashion suitable for the subject-oriented education tertiary students often experience. It calls for a different approach to education. With this approach 'a situation is presented before any knowledge is given. Then once the knowledge is acquired, it is applied back into the problem. The students are in control because they must select the knowledge needed to solve the problem, learn that knowledge and relate it to the problem' (Woods, 1987, p. 19).
The student-centred focus of problem-based learning and the iterative nature of learning through problem solving are an integral part of process-based learning. In Database 1 the two approaches also share the focus on solving real industry cases.
At Swinburne a pedagogy that has led to current forms of flexible delivery and is based on a constructivist, student centred learning approach that resources students in ways that best facilitate their learning, is known as Multi-Modal Learning (MML) (Markman, 1993; Jeffery, Smith & Weal, 1998). Like other resource-based learning systems (Noble, 1980) MML puts the information needed by students before them. The multiplicity of resources made available to students in Database 1 include a comprehensive Subject Learning Guide [a detailed description of the pedagogy behind the use of Learning Guides can be found in Jeffery, Smith & Weal (1998)], online support materials for every aspect of their on-campus learning (plus a CD of everything online for use off-campus), Video-on-Demand recordings of lectures and tutorial exercises, group assignments, face-to-face interaction with staff (lectures, tutorials and laboratory sessions), as well as a text book. Another resource is the use industrial case studies that guide the student through a series of comprehensive problems to learn how to engage in the process of data modelling for themselves.
Often resource-based learning is a part of distance education systems where students need information and academic staff are not readily available. Jevons and Northcott point out though that this tie to distance education is not explicit, that on- and off-campus study can be facilitated through the use of resource-based learning (National Board for Employment, Education and Training, 1994). Swinburne University of Technology in fact is not focussing on distance education, but is instead concerned to facilitate & complement on-campus study programs.
One reason for adopting a resource based learning pedagogy involves the issue of staffing the subject. There are really two aspects to the staffing issue.
The first staffing issue is consistency in approach. Database 1 has an annual enrolment of over 1,000 students. For instance in semester 1, 1998 there were 38 tutorials (tutes), 38 laboratory sessions (labs) and 4 lectures running every week. Students were encouraged to attend one of each, every week. More than 80% of tutes and labs were conducted by sessional staff. These sessional staff in the past have mostly been industry practitioners, but that is changing.
Non-traditional universities, such as Swinburne, are attempting to increase the research profile of the institution and its staff. Thus there is a marked increase in the numbers of post-graduate students studying at the university. These students are increasingly being employed to teach tutes and labs. Amongst the rationale given are that these students have the appropriate discipline expertise needed to teach undergraduate subjects, and given they are engaged in discipline specific studies, they are presumed to have enthusiasm for their subject matter. If then they teach in a related area, some of their enthusiasm will be communicated to their students.
Yet from an educational perspective such dependence on a transient workforce, often composed of people who have no training in nor commitment to teaching, can lead to a learning environment can be undesirable. A possible consequence is that teaching quality decreases. MML seeks to address these concerns and has been shown to improve the quality of student learning (Jeffery, Smith & Weal, 1998).
The second staffing issue is particular to the Database 1 subject. Conceptual modelling requires expert data modellers. Yet it is unrealistic to expect all sessional tutors to be such. There is then a need to appropriately resource not only the students, but also the staff teaching into the subject. These staffing issues give further impetus to the pedagogical viewpoint, that it is important to resource the students as well as possible with quality learning materials that to a significant extent are independent of the teacher face-to-face before them.
Even when staff are expert data modellers, they do not necessarily know how to best articulate data modelling strategies in ways that will most help their students. Again there is a need for training. By means of extensive experience experts build up cognitive maps (Cooper, 1998) that, in the case of data modelling, help them solve complex problems, without necessarily being aware of the formal methodology that is the basis of their work. Many of the sessionals hired for teaching in the subject likewise exhibit competence in database design, yet are not familiar with the conceptual modelling approach. This means that the learning material for the students can also be of help to the transient teaching staff to help them better understand their own expertise in the area & how to impart that to the student body. This is further necessitated by the fact that conceptual modelling in first databases courses is unconventional and therefore there are insufficient literary sources to support students at this level, let alone to support academics teaching conceptual modelling at this level.
Learning structures are important to the pedagogy in two ways. Firstly there is the structure of learning of the expert data modeller and secondly there is the need to integrate the whole subject learning into a cohesive whole.
The first of these is addressed in the way in which students are resourced with their MML materials. These deliberately include numerous worked examples (some textual and others via Video-on-Demand) that show the process by which an answer was derived, with possible deviant interpretations along the way. This is opposed to other first courses in database where students could simply be given a problem and a solution. However in process-based learning, an understanding of the process by which a solution might be derived is vitally important. This process is informed by how the expert data modeller might approach the task of solving a problem.
Experts in various fields build up cognitive schemata that represents their learning. In a sense then the student needs to study a number of real world case studies and go through the process of learning that an expert might. Cooper (1998, p. 1) says: "Only after sufficient schemas have been acquired will a learner be able to engage in mental rehearsal of instructional procedures that are accurate & meaningful." He goes on to say that studies investigating expert-novice differences in cognitive domains have shown that experts employ a hierarchical network of knowledge consisting of both procedural & declarative information. They have a library of mental constructs that enable them to recognise patterns as belonging to previously learned categories & which specify the types of responses appropriate to that category. It is these types of mental maps that students are encouraged to build through the highly visual process of conceptual modelling. Experts in fact are often not consciously aware of the mental processes at play in their work.
In part this was alluded to above in the discussion about staffing issues. Many of the sessionals employed to teach in Database 1 bring to the task a significant base of experience.
Hence they may employ conceptual modelling often without awareness of the formal methodologies involved. Yet the methods rather than their years of experience is what the students need to learn. The methods in combination with process-based learning on real world case studies facilitate student building of the mental maps they need to solve even unseen data modelling problems (the ultimate examples of which for them will be in their final examination). In Cooper's words (1998, p. 2): "The primary goal of education & training is to develop suitable instructional materials & activities which will enable novices to make the transition towards expertise."
The second aspect of learning structures is related to the first but applies to the subject as a whole. It is best depicted in the concept of the Learning Map.
A study conducted during the days when Database 1 catered predominantly to business students (Burmeister, 1995) it was found that many of the students seemed to have a myopic view of their study; they failed to take in the bigger picture of how a current topic of study fitted into the overall structure of what was being taught in the subject. To facilitate student understanding of the overall picture of the subject, i.e., to integrate the subject into a whole, teaching staff introduced a "learning map" which has evolved over the years & now is an integral part of the subject learning guide & the online education materials. It is on the front cover of the learning guide & accessible from every online module - in fact the portion relevant to the current topic being studied is reproduced online & if a student "clicks" on it they are taken to the whole learning map where they immediately see how what they are now engaged in fits into the overall scheme. Similarly, through the use of the learning map on the cover of the Subject Learning Guide students are provided an initial structure for the subject to facilitate an overall understanding of the content and processes being taught. This is done to facilitate the process for students of developing their own learning maps to structure their learning and build bridges between new and old material they had learnt. Yates and Chandler see learning maps as "instructional scaffolding" that facilitates construction of new knowledge. They say: "Contemporary cognitive theories of learning see the development of knowledge as a process of active construction" (Yates & Chandler, 1994, p 5). Research into learning (Evans, 1991; Lawson, 1991) shows that students who achieve the best results in the tertiary environment tend to be ones who are able to structure their learning, integrating concepts and thus are able to facilitate their recall of pertinent information and apply it in contextually relevant situations. Yates & Chandler (1994) who use synonymous terms such as "mental models" and "mental schemas" see the use of learning maps as an integral aspect of using prior knowledge to best advantage. The larger the knowledge base of the learner, the easier it becomes for them to acquire new knowledge in the same area.
Thus the learning structures employed in Database 1 attempt to extend the pedagogical relevance of cognitive schemata used to help students integrate all aspects of the subject, to helping them build mental maps that facilitate better data modelling.
At Swinburne, the tradition of vocational education, the reality of a student body with diverse interests and backgrounds, the economic pressures and well as the professional demands of the workplace constitute a set of opportunities and constraints that bear on the curriculum in database analysis and design. We propose a solution to this problem that we assert is better than the traditional approach to curriculum in this topic area. The proposed solution is to apply in significant measure the well understood yet still unconventional techniques of conceptual data modelling. The notations used are the conventional Entity-Relationship and Functional Dependency diagramming. The delivery of this content poses certain challenges, but experience shows that it is feasible with a Multi-Modal, process-based approach well resourced with authentic case studies.
ACM/IEEE-CS Joint Curriculum Task Force (1991), "Computing curricula 1991", Allen B. Tucker et al. (editor and co-chair), ACM Press : IEEE Computer Society Press
Atzeni P. & Parker D.S. (1982) "Assumptions in Relational Database Theory", Proc. SIGACT-SIGMOD PODS, 1-9
Australian Committee for Training Curriculum (1994), "National Computing & Information Technology Curriculum", ACTRAC Products Ltd
Barker, R. (1990) "CASE*METHOD Entity Relationship Modelling" Addison-Wesley
Batini C., Ceri S., Navathe S. B. (1992), "Conceptual Database Design: An Entity-Relationship Approach", Benjamin Cummings
Borgida A. (1991) "Knowledge Representation, Semantic Modelling: Similarities and Differences", in Entity-Relationship Approach: The Core of Conceptual Modelling, H. Kangassalo (Ed), Elsevier
Burmeister, O.K. (1995) "Evaluating the factors that facilitate a deep understanding of data analysis", Australian Journal of Information Systems, Vol. 3, No. 1, Sep.
Carter J. (1995), "The Relational Database", Chapman & Hall
Chen P.P. (1976), "The Entity-Relationship Model - Towards a Unified View of Data", ACM TODS, Vol 1, No 1, 9-36
Chen P.P. (1983) "English Sentence Structure and Entity-Relationship Diagrams", Information Sciences 29(2), 127-149
Codd, E. F. (1970) A relational model for large shared databanks. Comm ACM Vol.13 No.6, 377-387
Cooper, G. (1998) The Application of Mental Rehearsal to Cognitive Domains, AARA Conference paper, http://www.swin.edu.au/arre/98pap/coo98034.html.
Date C.J. (1986) "A Practical Approach to Database Design", Chapter 19 of "Relational Database Selected Writings", Addison Wesley
Dutka, Alan F. & Howard H. Hanson (1989) "Fundamentals of data normalization", Addison-Wesley,
Eden P. (1996) A Step by Step Method for Conceptual Data Analysis, Software Engineering: Education and Practice, IEEE Computer Society Press, 42-49
Eden P. (1997), "Entity Relationship Analysis", Swinburne Press
Edmond, D. (1992) Information Modelling, Prentice-Hall
Elmasri R., Navathe S.B. (1989), "Fundamentals of Database Systems", Benjamin Cummings
Evans, G. (1991). "Student control over learning", Teaching for Learning, Hawthorn: ACER, pp 126-145.
Farouque, F. (1994) Older unis don't promise a job: study, The Age, 25 May 1994.
Halpin T.A. (1995) "Conceptual Schema and Relational Database Design" Prentice-Hall
International Standards Organization (1987), "Information Processing Systems - Concepts and terminology for the conceptual schema and the information base", TR 9007
International Standards Organization (1991), "Information Retrieval, Transfer and Management for OSI: The General Relationship Model - Working Draft", ISO/IEC JTC1/SC21 N4975
James, J. A. & Olson, N. (1994) Fact-based Corporate Data Modeling, First International Object-Role Modeling Conference; Magnetic Island, Queensland, pp. 755-775
Jeffery, P.L., Smith, R.N. & Weal, S.E. (1998) The Role of Learning Guides in Multi-Modal Learning at the Lilydale Campus of Swinburne University of Technology, AARA Conference paper, http://www.swin.edu.au/arre/98pap/jef98078.html.
Kent, W. (1983) "A Simple Guide to Five Normal Forms in Relational Database Theory", Comm ACM, 26 (2), 120-125
Lawson, M. (1991) "Managing Problem-Solving", Teaching for Learning, Hawthorn: ACER, pp 126-145.
Loucopoulos P. & Theodoulidis B. (1992), "CASE - Methods and Support Tools" in Loucopoulos & Zicari (1992)
Loucopoulos P. & Zicari R. (1992) "Conceptual Modelling, Databases and CASE", Wiley
Markman, P. (1993) "Multi Modal Learning", 25 March (A guide for staff developing multi modal learning materials).
McAdam, J. (1999) Swinburne University of Technology 1999 Marketing Plan, Hawthorn: Swinburne Press.
McFadden Fred R. & Hoffer Jeffrey A. (1994) Modern database management 4th ed. Redwood City, Calif : Benjamin/Cummings
National Board for Employment, Education and Training (1994) Cost and Quality in Resource-Based Learning On- and Off-Campus, Commissioned Report No. 33 (Jevons/Northcott), AGPS: Canberra, October.
Noble, P. (1980) Resource-Based Learning in Post Compulsory Education, Kogan Page: London.
Peckham J. & Maryanski F. (1988) Semantic Data Models. ACM Computing Surveys, Vol.20, No.3. 153-189
Pristel, S. (1994) Swinburne produces jobs faster, The Herald Sun, 25 May 1994.
Richards, C. (1995) New unis score in job stakes, The AGE, 29 November 1995, p. 6.
Rock-Evans R. (1989), "An Introduction to Data and Activity Analysis", QED Information Sciences
Simsion G. (1994) "Data Modelling Essentials", Van Nostrand Reinhold,
Tsichritzis D.C. & Klug A. (1978) The ANSI/X3/SPARC Framework, AIPS Press, Montvale, N.J.
Wand Y. & Weber R. (1989) "An ontological analysis of systems analysis and design methods", in Information Systems Concepts - An In-Depth Analysis, E. Falkenberg & P. Lindgreen, Eds., North Holland, 79-107
Winograd T. & Flores F. (1986) Understanding Computers and Cognition, Addison-Wesley Publishing
Woods, D. (1987) "Problem-based learning and problem-solving", Problem-Based Learning in Education for the Professions, Ed. D. Boud, HERDSA: Sydney, pp 19-42.
Yates, G. & Chandler, M. (1994) "Prior Knowledge", SET: Research Information for Teachers, No 2, Item 6.
Yourdon, E. (1989), "Modern Structured Analysis", Yourdon Press