Conceptual Workflow Modelling for Remote Courses José ... - CiteSeerX

3 downloads 82434 Views 83KB Size Report
Proceedings of IFP World Computer Congress, Teleteaching '98 .... the whole process, with the clear definition of all the activities to be executed, of their.
Proceedings of IFP World Computer Congress, Teleteaching '98 Distance Learning, Training and Education, Proceedings of the XV World Computer Congress, IFIP, 31 de agosto a 4 de setembro de 1998, Viena - Aústria e Budapeste - Hungia, p. 789-797.

Conceptual Workflow Modelling for Remote Courses * José Palazzo M. de Oliveira1 Mariano Nicolao1,2 Nina Edelweiss1 1

Universidade Federal do Rio Grande do Sul Instituto de Informática Av. Bento Gonçalves, 9500 Caixa Postal 15064 91501-970 - Porto Alegre - RS - Brazil e-mail: [email protected], [email protected]

2

Universidade Luterana do Brasil Departamento de Informática Rua Miguel Tostes, 101 Caixa Postal 124 92420-280 Canoas - RS - Brasil e-mail: [email protected]

Abstract Development of a wide spread project intended to teaching Computer Science, integrating a considerable number of students all over a country with big geographical extension and scarce educational funding such as Brazil, is a complex task. The communication technology is an important factor in this case to maintain the expenses under control and assure the participation of the geographically dispersed communities. The new tendency to deploy the learning activities to the WWW creates an excellent opportunity for remote courses. To establish a god quality in the offered learning material, the remote (and Web) technology demands new information system modelling techniques. Our approach is to develop a complete conceptual modelling of the course. The static parcel is described with an entity-relationship model and the dynamics properties with a workflow modelling of the full course structure and temporal constraints. To make the traditional workflow modelling more effective it is necessary to improve the conceptual level specification. The unified model, here described, is able to represent the course internal behaviour and the relationship with the environment. In this work, a technique of conceptual modelling of workflow is presented using an object-oriented temporal model, the TF-ORM. The main objective of this paper is to show as a formal entity-relationship and a workflow description can be used to generate the conceptual schema of the course and the set of rules for the operational management. Additionally a graphical representation is introduced to allow visualisation and validation by the users.

1. Introduction One of the most innovative approaches in the use of the Internet and of computers is the application of a new class of self-study. This is a revolution in the teaching style. The old theories and practices, based in the traditional book and paper style of classes, must be reviewed [12]. Some of the efforts in the use of the WWW in education may be found at

*

This work was partially supported by CNPq

the “The World-Wide Web in Education” [14]. A substantial amount of work in multiple areas is under development, a comprehensive index may be found at “The 13 Member's Courses” [13]. Teachers are now developing web pages, intelligent agents and interactive courses based on the new technologies. The static and predefined classes structure is being replaced by flexible software capable to offer an individual path, tailored to each student personal skill. This approach is well suited to accomplish a free autonomous continuing education or to support the traditional classroom teaching. One proposal of the first category is the Electronic Education Environment, or E3, developed at the University of Austin, Texas, to support process in a Virtual University (VU). A VU focuses on developing skills and expertise by mass customising content on demand rather than providing terminal degree programs with homogeneous and predetermined curricula [2]. The second category is exemplified by [9] where virtual classes are supported via an extended browser and traditional videotape libraries. Other example of this category of course is the database course developed at Politecnico of Milan [1] where the courses are stored and the web pages generated on-the-fly from a lectures' database. The main objective of this work is the modelling of the Web contents of the course. Our work, partially described in this paper, is centred in the dynamic representation associated with the implementation of a semi-automatic workflow support intended for the implementation of a course on the Web. In this paper, the course is modelled in both the static and dynamic aspects. The static representation corresponds to the modelling of the course structure. The dynamic representation is developed using workflow modelling and corresponds to the detailed student activities. This paper is organised as follows. In section 2 the full modelling of a remote teaching course is presented, in Section 3 the formal model is shortly presented. Section 4 analyses the construction of a workflow model and the modelling technique is detailed, finally, in Section 5 a short description of a lecture model is presented.

2. The course modelling Using the Web as an enabling technology to support a traditional degree program requires a formal following and evaluation of the student individual effort and results in order to obtain a rigorous evaluation. In this case, a more rigid structure should be imposed to the students with specific subjects lectures and exercises resolution planed for each period. The teaching work is composed by different activities, executed by students, human tutors and automated processes in a (partially) defined sequence. As the course become more complex, the co-ordination of the execution becomes an important feature to be considered while planning activities. The representation of the programmed courseware activities prior to the implementation becomes, therefore, of fundamental importance. The definition of the whole process, with the clear definition of all the activities to be executed, of their relationships, and including the agents responsible of their execution, is known as workflow. One of the main goals of workflow modelling is to decrease the number of problems due to activities’ co-ordination. In traditional courses, usually it is not possible to have detailed control of all the activities to be executed. This important aspect must be clarified by the workflow modelling process. The workflow model shall define not only the activities, but also most of the temporal restrictions to their execution, the dynamic data and control exchange between activities, and the persons (agents) responsible for each activity [7]. Temporal aspects are of fundamental importance in courseware. Temporal rules control the synchronism of the different tasks and activities involved in the course development.

The construction of a course workflow model allows during the construction of the model the whole process is analysed, in terms of activities to be executed, and the sequence of their execution. This is done in an abstract way with the aim of solving detected inconsistencies [8] and, also, to achieve a better work distribution along the course.

Author

Produces

Course

Composed by

Has

Elements Inscribed

Temporal constrains Student

Exercises

Texts

Links

Controls

Expert system (agent)

Reports to

Tutor

Student data

Executes Proposes Works with

Completed exercises

Monitors

Evaluates

Has

Figure 1: The course model

Our project in remote teaching is based in the course conceptual modelling as a workflow and in the automated agent, following the course specification and monitoring each of the students. The static model is represented as an entity-relationship schema (Fig. 1) and the dynamic structure is modelled as a workflow (Fig. 3). The course is described, in the E-R model, as relationships among the main human entities Author, Student, Tutor and the automated entity Agent. The Author produces a course writing and defining the Elements that may be Texts, WWW links to be browsed and, finally, Exercises to verify the concepts acquired by the students. An important characteristic in remote courses is the necessity of the temporal constraints definition of the activities. The students may have some temporal limits for the execution of a specific activity, as the restriction to complete a course in a three-month period. In the model, these constraints are associated with the course as the full set of temporal constraints are interrelated. Each inscribed student has a personal set of data logging all the activities performed, the exercises results and the associated temporal information. The human Tutor follows the Student activities using this data. As we intent to allow a large number of students to follow the courses, it is onerous to offer a continuous attention to the students with a large number of human tutors available 24 hours. The solution is to provided by an automated

Agent that not only monitors the student activities but also advise he or she about the preferred lectures sequence, available links to complement the studies and exercises to be executed. Based on the exercises results, the agent may suggests more lectures or proposes new links as complimentary material to help in the errors correction. To achieve this kind of reactive comportment in a semi-automated system it is imperative to develop a complete and dynamic course description. A big effort is being developed in the course modelling using an object-oriented temporal model intended to represent the workflow. This modelling is presented in the next sections. The Agent is active element, an expert system, that based in the information coded in the workflow model interprets the students activities and offer an semi-automated support. In the cases where the Agent is unable to advise the student, an exception report is send to the Tutor that will contact directly the student.

3. TF-ORM TF-ORM (Temporal Functionality in Objects with Roles Model) [4, 5] is a temporal object-oriented data model. It differs from other temporal object-oriented data models by the use of the role concept to represent the different behaviours of an object. The role concept associated to the object-oriented paradigm was introduced with the ORM model (Objects with Role Model) [11]. A class presents different roles and these can be dynamically instantiated. Through the role concept, the dynamic aspects of an object are separated from the static ones. An object is still an instance of only one class, but it can play different roles during its lifetime. An instance of an object can play different roles during its lifetime. In addition, an object can have more than one instance of the same role at the same time. Three different kinds of classes are identified: process classes, resource classes and agent classes. They are all modelled in a similar way. A class is defined by a unique name and a set of roles: classi = (cni, R0, R1, ... Rn)

Each role is defined by a name (rni), a set of properties (Pi), a set of abstract states the role can assume while playing this role (Si), a set of messages the role can receive or send (Mi), and a set of rules - state transition rules and integrity rules (Rui): Ri = (rni, Pi, Si, Mi, Rui)

Properties may be static (having the same value all over the instances lifetime) or dynamic (when they may assume different values with time). Dynamic properties have two different time points associated with each value: the transaction time, corresponding to the moment when the information is introduced in the database, and the valid time, the time when that information starts to be valid in the real world. The messages represent the objects’ methods. Incoming and outgoing messages are defined, and the class sending or receiving the message shall be informed. Values to be used in the methods corresponding to the messages are passed by the way of message parameters. Human decisions are represented in agent classes as an incoming message. The state transition rules define the dynamic evolution of an object. The arrival of a message sent by another class does not mean that the corresponding method will always be executed - state transition rules control these messages. Such a rule defines a combination of an object state s1 and an incoming message mi1 to change to another state s2. One or more messages can be send when a transition rule is executed (mo1 through mon). A transition condition can be associated to a rule acting as conditions that constrain the state transition - the transition will only occur if this condition is true. This condition is represented by a logic formula. With this construction, temporal integrity conditions can be

represented and invalid transitions are not executed. The general form of a state transition rule ri is the following: ri: state(s1), msg(mi1) ⇒ msg(mo1), msg(mo2), ..., msg(mon),state(s2); []

msg(m1)

activity 1

msg(m2)

msg(m1)

activity 2

msg(m2)

activity 1

msg(mn)

st(s1)

.. .

st(s1) activity 1

msg(mo)

activity 2

activity 1

st(s2)

st(s2)

(a) sequential

(b) total convergent

msg(m1) msg(m2)

.. .

msg(mn)

msg(mi)

k=2

msg(mj)

activity 1

msg(mo)

activity 2

st(s1) activity 1

st(s2)

(c) partial convergent msg(m1)

activity 1

activity 3 msg(m3)

msg(m2) activity

.. .

st(s1)

2

msg(mn)

msg(m1)

activity 5

activity 1

msg(m2)

msg(m5)

st(s1) activity 1

activity with transition condition

activity 1

st(s2)

st(s2)

(d) parallel synchronisation (e) Conditional Figure 2: Graphical representation of synchronisation

A rule may also be defined based on the arrival of a set of messages - the rule is only executed when all the messages have arrived, independently of their order or arriving time. This is represented as follows: ri: state(s1), {msg(mi1), msg(mi2)... , msg(min)} ⇒ msg((mo1), state(s2); []

Constraint rules are represented by two conditions: if the first condition holds, the second shall also hold. Temporal conditions can be used in both forms of rules. The set of rules completes the object’s behaviour definition. Three types of classes can be defined: (i) resource classes, modelling information and resources; (ii) process classes, representing the processes to be executed with this

information and the resources; and (iii) agent classes, representing the persons that carry out the processes. In addition to the above-mentioned messages, agent classes also include human decisions representing the non-structured work in the formal definition environment. Temporal information are associated to all the instances (class and role instances) - the instance's creation time and destruction time, and the time instants in which the instance's activity was suspended and resumed. The temporal information is stored in special predefined properties and can be used by the query language [6] to retrieve information. Predefined properties are inherited from a superclass Object, from which all the TF-ORM defined classes are sub-classes. Each class presents a special role, the base role, where the global properties inherited by all other roles and the initial characteristics of the other roles are described. The TF-ORM model supports specialisation and aggregation mechanisms, with the possibility of inheriting roles, or redefining them.

4. Workflow Modelling The workflow modelling using the TF-ORM is accomplished by the definition of three types of classes: agent class, recourse class and process class. Most of the existent models of workflow do not represent the work portion with human intervention (unstructured work) that is necessary to represent a course. TF-ORM allows the representation of the unstructured work portion of the processes through agent class. An agent class represents people acting in the system. The agents have an own functionality that is the human decision as the selection of one of a list of links to be followed or the choice of alternative texts. A decision represents the result of a formally undefined process made by a human agent. The resource structures (data, documents) are represented by the resource classes. The process classes integrate these agents and resources describing the work organisation developed in the application and the co-operation among the agents. Messages describe interactions among activities, determining the control flow and allowing the synchronisation of those activities execution process. The possibility of formal representation of these interactions is a fundamental factor that must be present in a workflow model. The activities execution order (synchronism) determine the evolution of the work in a workflow. This synchronism is represented in TF-ORM through states transition rules. Through state transition rules the following synchronism conditions can be represented: sequential, convergent (join), divergent (fork) and conditional. Sequential: in this case, the activities are scheduled in sequential form, obeying a fixed execution order. When the activity, is in the state s1, receives the message m1, it sends the message m2 and changes to state s2 (fig. 2a). The syntax in TF-ORM is: st(s1), msg(m1) ⇒ msg(m2), st(s2); Total convergent: this situation represents an activity being fired in function of events determined by one or more other activities. When the activity receives a group of messages {m1 ... mn} in the state s1, then this activity sends the message mo and changes of state, assuming the state s2 (fig. 2b). The arrival order of the messages is irrelevant. The syntax is: st(s1), {msg(m1), msg(m2), ..., msg(mn)} ⇒ msg(mo), st(s2); Partial convergent: for situations in which n messages can arrive but just k messages are required to fire the transition, we defined the operator k {msg(m1), msg(m2), msg(mn)}, where k (1 ≥ k < n) determines the number of necessary messages to fire the transition. In addition, in this case the arrival order of the messages is irrelevant (fig. 2c). A syntax example and the graphical representation is shown below: st(s1), 2{msg(m1), msg(m2), ..., msg(mn)} ⇒ msg(mo), st(s2);

Parallel synchronisation: this corresponds to the situation in which different activities can be fired simultaneously. If the activity is in the state s1, then this activity sends the messages m1 and m2 changing of state, assuming the state s2 (fig. 2d). The representation through TF-ORM is: st(s1) ⇒ msg(m1), msg(mn), st(s2); Conditional: an activity will be fired in function of a condition to be satisfied in this situation. If the instance receives the message m1, in the state s1, and if the transition condition is satisfied, then this activity sends the message m3 changing of state, assuming the state s2. In the same way, if the activity receives the message m1, in the state s1 and if the transition condition is satisfied, then the activity sends the message m5 assuming the state s2. The syntactic and the graphical representation are exemplified below: st(s1), msg(m1) ⇒ msg(m3), st(s2); () st(s1), msg(m2) ⇒ msg(m5), st(s2); () The state transition condition is optional and it may be applied for all state transition rules. The workflow models need to support expressions related to the processes, temporal restrictions, dynamic changes and treatment of exceptions [7] what not always happens in the proposed modelling techniques. Temporal Modelling allows the representation of many of the aspects mentioned above. Through these techniques the dynamic characteristics of the applications and the temporal interaction are represented among different processes. The possibility of storing, manipulating and recovering temporal data should be considered when choosing a method of workflow modelling. A course modelling needs time restrictions control. These controls produce warnings (eg.: First Reminder After x days, Repeat Reminders Every y days, Deadline Occur After z days) automatically to alert the students and the tutors of pending activities, as well as to fire automated activities related with time. In TF-ORM these controls are represented by the states transition rules, more specifically in the transition condition. The conditions are expressed through sentences of a first order temporal language [3, 5]. In these conditions values of properties (presents, past, futures) and states of roles can be evaluated. The role identifiers can also be referenced as they are stored in the pre-defined properties oid and rid. Some operators and time functions are: sometime past, immediately past, always past, sometime future, immediately future, always future, since, until, before, after The temporal functions: value, past_value, duration, now, valid_time, transaction_time, role_creation_time, class_creation_time, class_end_time, state, state_at, year, month, day, hour, minute, weekday, lower_bound, upper_bound.

5. The dynamic model This model, besides being temporal and using the object-orientation paradigm, uses the concept of roles to represent different behaviours of objects, allowing a richer representation of the workflow. The workflow tasks are represented as classes, encapsulating the representation of the activities of each class. The definition of all the classes’ interfaces with other tasks, and with agents and resources involved in the workflow completes the modelling. The processes involved in a workflow, their relationships and co-ordination are defined, but also the data flow between these processes and the identification of agents to execute the processes is accomplished. In addition, a set of temporal logic rules incorporate a solid formalism to express reactive computations, usually influenced by events external to the workflow, like exceptions and pre and postconditions associated to the processes execution. The final model is a formal model, and

can be used to analyse the workflow, identifying possible definition problems that can be solved prior to the implementation of the course workflow.

Links

Fig. 2d

Nnn nnn n Nnn nnn n

Following lecture 1

...

Nnn nnn n

Fig. 2a Reading text 1.1

Browsing link 1.1

Browsing link 1.2

...

Browsing link 1.n

... Reading text 1.2

k=2 Fig. 2c Reading text 1.3

Executing exercise 1.1

Fig. 2b

Figure 3: The dynamic workflow model

The automatic supervision made by the Agent is controlled by a detailed course specification in the formal workflow specification. A partial example is presented in Fig. 3, for sake of the simplicity this is a greatly simplified schema. A Student when inscribed in the course is at the state "Folowing lecture" this state sends two messages, one starts a new role "Reading text 1.1" and changes the state to active. The second fires a human decision, represented in the figure by the form Links, the student may select one from various links associated with this lecture. The browsing of the links is optional but it is required, at the end of the lecture, that the student has worked with at least two links (k =2). After the completion of the reading of the three texts and after the browsing of a minimum of two link, messages are send changing the state to "Executing exercise 1.1". With this formalism, it is possible to represent in detail the course structure. With a stored representation, it is possible to manage the execution of the individual students tracks in a semi-automated way. The use of a powerful data model such as TFORM does not require a specific database management system to support it to implement

an application. Existent commercial DBMSs can be used for this purpose as long as a proper mapping from the temporal data model to the data model underlying the adopted DBMS is provided. Since these systems usually do not provide a built-in support for temporal features, this mapping requires some transformations and the translation of the temporal schema and the temporal queries to the language of the adopted DBMS [10].

6. Conclusion In this paper, a technique of conceptual modelling of workflow was presented using the model TF-ORM. This technique is intended to specify and to support the workflow implementation of a remote course. In this model, syntactic constructions are introduced to obtain the specification modularization and workflow parallelism in an efficient way. The main objective of this paper is to show how a formal entity-relationship and workflow description can be used to generate the complete model representing an infrastructure to support a distributed course environment. Additionally a graphical representation was introduced to allow visualisation and validation by the users. In addition, the transition rules provide a convenient formalism to express reactive computations as they are influenced by external events, generated out of the Workflow Management System, as exceptions or non anticipated pre-conditions of tasks.

References [1] F. Casati, B.Pernici. The Design of Distance Education Applications based on the WWW , Web Based Instruction, edited by Badrul Khan, Published By Educational Technology Publications, Sept. 1996 [2] R. Chellappa, A. Barua & A.B. Whinston, An Electronic Infrastructure for a Virtual University, Communications of the ACM, Sep. 97, v. 40, n. 9, p. 56-58. [3] N. Edelweiss, B. Pernici, J. Palazzo M. de Oliveira, J. M. V. de Castilho, Extending an Object Oriented Model to Represent Temporal Requirements, Proceedings, 12th International Conference of the Chilean Computer Science Society, Santiago, Chile 14 a 16 de outubro de 1992, p. 11-25. [4] Edelweiss, N., Oliveira, J.P.M. and Pernici, B., An Temporal Object-Oriented Model, Proceedings, 5th International Conference on Advanced Information Systems Engineering, Paris, France, June 8-11, Lecture Notices in Computer Science n. 685, pp.397-415. [5] Edelweiss, N., Oliveira, J.P.M. and Castilho, J.M.V, Temporal Logic Language for Temporal Conditions Definition, Proceedings, 13th International Conference of the Chilean Computer Science Society, La Serena, Chile, Oct. 14-16, pp.163-178. [6] Edelweiss, N., Oliveira, J.P.M. and Pernici, B., An Object-oriented approach to temporal query language, Proceedings, 5th Database and Expert Systems Applications Conference, Athens, Greece, Sept. 7-9, Lecture Notes in Computer Science n. 856, pp.225-235. [7] ELLIS, Clarence A. and Nutt, Gary J. - Workflow: The Process Spectrum, Departament of Computer Science - University of Colorado - Boulder, CO 80309-0430. [8] Georgakoupoulos, D.; Hornick, M.; Sheth, A. An Overview of Workflow Management: from process modeling to workflow automation infrastructure. ACM Distributed and Parallel Databases, n.3, p.119-153, Sea. 1995. [9] Johnson, W.L. and Shaw, E., Using Agents to Overcome Difficulties in Web-Based Courseware , To be presented at the AI-ED'97 Workshop on Intelligent Educational Systems on the World Wide Web, August, 1997. [10] Implementation of an Object-Oriented Temporal Model, J. Palazzo M. de Oliveira et. al.; Proceedings of the DEXA ‘95 Conference Workshop, London, Sep. 4-8, 1995, p. 35-44. [11] Pernici, B., 1990, Objects with Roles, Proceedings, ACM/IEEE Conference on Office Information Systems, Cambridge, MA, April 25-27, SIGOIS Bulletin, v.11, n.2-3, pp.205-15. [12] D. Spender, Revolution in Education Discussed by Keynote Speaker at IFIP Congress ’96, IFIP Newsletter, V. 14, n. 1, March 97, p. 1,9,10 – the full spech may be found at http://www.acs.arg.au/ifip96.html. [13] http://www.unb.ca/web/13/mandate.html [14] http://tecfa.unige.ch/tecfa/tecfa-research/CMC/andrea95/andrea.html