A framework for the specification of the semantics and the dynamics of Instructional Applications.

An instructional application consists of a set of resources and activities to implement interacting, interrelated, and structured experiences oriented towards achieving specific educational objectives. The development of computer-based instructional applications has to follow a well defined process, so models for computer-based instructional applications have been proposed. Some of these models are oriented towards implementing products delivered in a rather specific technical platform. But specifying instructional applications using generic formats that are independent from the implementation platform, as Educational Modelling Languages do, makes it possible to improve their interoperability as well as their reusability.

The work presented here proposes a specification framework based on two formalization levels. A first level proposes a meta-model to specify an instructional application by means of two models to represent the application from two different perspectives: the semantic model gathering its structure and the instructional relationships existing among its components; and the operational model to establish the application operation and behaviour. The second level of the framework implements the application model, which uses XML technologies to represent and process its components, relationships, and behaviour. A tool has been developed to visualize the specification of the model entities as well as to check some properties of the instructional application.

**********

An Instructional Application (IA) consists of a set of resources and activities to implement interacting, interrelated, and structured experiences oriented towards achieving specific educational objectives (Nervig, 1990). Examples of IA are not just limited to traditional courses; they may include many kinds of learning experiences such as topic introduction, problem-solving, specific tutorial and training processes, or the use of expert systems. IAs have taken advantage of computers and other information technologies to provide access in a distributed and friendly way. Thus, several techniques have been used to develop and improve the learning and teaching processes, including Artificial Intelligence, to adapt the system to the user needs and to manage the knowledge to be taught (e.g., Intelligent Tutoring Systems, Burns & Capps, 1988), telecommunication networks and web technologies to support the access to educational resources (e.g., web-based courses that can also be adapted to the user features, Brusilovsky, 1998) or the application of Human Computer Interaction (HCI) principles and usability engineering to improve the interaction with the IA (Landoni & Diaz, 2003).

Nevertheless, pure instructional aspects are somehow undervalued in the computer-based IA development where technology is often considered as the main subject instead of being considered a means to reach certain educational objectives. Indeed, most of the current development tools address the production of IA in such a way as to be deployed in "online" e-learning environments and little attention is given to the previous development steps, specifically, to the specification of instructional issues. Indeed, relevant development activities such as the analysis of user goals and needs, or the design of the instructional tasks, and the educational contents that will support an effective and efficient learning process, are hardly considered except for some automated instructional design tools, such those reviewed in (Kasowitz, 1998), whose goal is to help designers in integrating instructional aspects into a computer-based IA. Thus, tools such as Instructional Design-Ware (Langevin Learning Services, 2002) or Designers' Edge (Mentergy, 2002) are addressed to "guide non-ID professionals through the process of creating effective instruction." MISA goes a step further (Paquette, Rosca, De la Teja, Leonard, & Lundgren-Cayrol, 2001) proposing an engineering method to develop IAs. The method consisted of several phases and the resulting products take into consideration the learning and user requirements.

Another problem arose due to the fact that most development tools are usually bound up with specific delivery technologies (e.g., HTML or PDF formats) or even with proprietary multimedia authoring tools. The production of generic instructional specifications that are platform-independent and, therefore, can be implemented and accessed in different environments, improves reusability and interoperability. This is the goal of Educational Modeling Languages (EML), which gather both instructional and technical aspects (European Committee for Standardization/International Society Standardization System [Cen/Isss], 2002).

The current work proposes a framework to support the specification of instructional issues in a computer technology context, not biased by their delivery formats and according to the line of other EML proposals. The specification framework introduced here is based on two formalization levels (Buendia & Diaz, 2002). The first level consists of a meta-model defined through UML notations (Booch, Jacobson, & Rumbaugh, 1998) that is oriented towards describing an IA from two different perspectives: the semantic model that gathers its structure and the instructional relationships existing among its components; and the operational model that establishes the application operation and behavior considering scenarios for different kinds of users, such as tutors, pedagogues or students. The second level is based on XML notations to specify the model entities, relationships and behavior in a format that enables their computing processing. In this article, we only focus on the first level. The main contribution of this framework, regarding to other EML proposals, can be summarized as:

* The proposed IA meta-model makes possible to specify the application from two different but complementary perspectives: a semantic model represents static aspects (both, components and relationships among them) and an operational model of the application will deal with dynamic issues.

* The relationships among the three main model entities (User profiles, Learning scenarios and Didactic structures) are set up in a dynamic way instead of being defined in a static and declarative way. This approach increases the IA flexibility as well as its adaptation to the user needs and preferences. For example, the best learning scenario for a given user or the best didactic structure for such scenario can be decided at runtime according to some adaptation rules that will determine the dynamic binding of the model entities.

* The use of educational ontologies, such as those proposed in (Mizoguchi, Sinitsa, & Ikeda, 1996) and (Leidig, 2001), that provide a formal vocabulary for identifying instructional issues and methods for managing them. This feature helps to configure relationships between educational goals and learning activities, or instructional and knowledge objects.

The remaining sections of this article are organized as follows. The second section introduces the proposed Instructional Application (IA) model and describes it from two different points of view: static and dynamic perspectives. The former used for the semantic model and the latter for the operational model. The third section reviews some existing EMLs and discusses the main contributions of our proposal. Finally, the last section presents some concluding remarks.

AN INSTRUCTIONAL APPLICATION MODEL

The meta-model presented in this article is oriented towards offering instructional designers a framework for the specification of any IA from two different perspectives:

* the semantic view, where the instructional aspects are specified using an information model representing the composition of learning activities, contents and user information as well as their relationships (whether static or procedurally defined); and,

* the operational view, where the application use and interactive behavior are defined for different kinds of users.

The next subsections describe this model from those perspectives: the semantic model will be represented by means of UML Class and Object diagrams; meanwhile Use Case and Sequence diagrams are deployed to describe the operational model (see UML summary in Appendix A).

The Semantic Model

The Instructional Application (IA) is composed by three main entities (Figure 1):

* User Profile (UP): stores relevant information concerning the learning and teaching processes for different kinds of users.

* Learning Scenario (LS): defined as the set of terms, conditions, and activities that characterize the user learning in a specific context.

* Didactic Structure (DS): used to organize educational resources and contents in a didactic way.

[FIGURE 1 OMITTED]

The next paragraphs describe the Model entities as well as their relationships.

User profile. User Profile (UP) entities store relevant information concerning the learning and teaching process. Figure 2 shows the UML diagram that displays the UP structure.

[FIGURE 2 OMITTED]

Each UP entity has some basic attributes such as the "id" that identifies the user whose profile is being defined and the "group" the user belongs to, since learning and teaching are considered as cooperative activities.

The user types are derived from the actors defined in (Lindner, 2001):

* Learner: as the user engaged in acquiring knowledge or skills.

* Author: as the user who develops contents and educational resources.

* Tutor: as the user who supervises Learners.

* Teacher: as the user who teaches about a specific knowledge domain.

* Pedagogue: as the user who performs the pedagogic guidance.

As it can be seen, staff users have been divided into several categories (Author, Tutor, Teacher, Pedagogue) emphasizing the different roles they play in the teaching process. For example, the Teacher organizes learning scenarios concerning a specific knowledge domain while the Pedagogue designs generic learning scenarios and strategies that will be used in different domains. Learner users have their own Learning Features, which represents the particular information related to their learning process. This information is structured in Preferences (e.g., the learning style), Objectives (e.g., an objective based on getting some technical skills), and Competency items (e.g., an exploring ability).

Moreover, a UP is composed by the following elements:

* Learning Management, representing the information related to the user interaction with LS entities (see description below). This information should be differently interpreted for each role, so for a teacher, the Active Learning Scenarios will refer to those specific scenarios on which he/she is directly involved meanwhile for a pedagogue, these entities will point to all learning scenarios since he is an expert in educational technology.

* Resource Management, which represents the information about the used resources used in the learning process. There are two kinds of resources: the Didactic Structure entities that are managed by Teacher users and the contents developed by Author users, stored on Educational Content Provider (ECP) repositories.

Figure 3 shows an example of UP entity that represents a Learner user who is identified as "aperez" and does not belong to any group. Learning Features includes two Preferences (about the user "Learning style" and his preferred "Interaction" mode), meanwhile an Objective item has been stated ("Getting technical skills"). The model also shows that this user has acquired a "Basic explorer" Competency as a consequence of finishing a "Computer-Tutorial" Learning Scenario and at this time, he is currently involved in "Computer-Training" Learning Scenario. As it can be seen the UP entity can be used to hold different kinds of useful information about the user and his status in the IA operation. This information might be used to control the learning and teaching processes although this is not a concern of the framework proposed here.

[FIGURE 3 OMITTED]

Learning scenario. A Learning Scenario (LS) is defined as the set of terms, conditions, and activities that characterize the learning process in a specific context. For example, an introductory lesson is a typical scenario that has a number of features (e.g., educational goals, structures, activities, etc.) that apply in different domains of knowledge. The purpose of the LS is to collect similarities that are context-independent into a unique entity that can be easily reused in different learning situations by adding the corresponding context-dependent information.

Thus, an LS entity is composed by three main elements (see the UML diagram in Figure 4) in such a way that there are no entities with the same values for the three elements:

* GoalStructure: represents the learning purposes of the LS entity, which are defined in a hierarchical way.

* ActivitySet: represents the actions and processes allowed to be performed in such a scenario.

* Management: describes the information to control the scenario operation as well as its outcomes.

[FIGURE 4 OMITTED]

The GoalsStructure element is organized by means of a hierarchical structure of Goal_Item elements. This hierarchy classifies Goal_Item entities into "Dependent-Domain" or "Independent-Domain" goals, each of which can be further divided into other categories such as "Memory" or "Reasoning," and so on. The lowest level is considered as the more specific category. Educational ontologies, such as the one proposed in (Mizoguchi et al., 1996), are used to define the Goal_Item contents.

The ActivitySet element has one or more Activity entities which represent generic learning processes that can be carried out in the scenario (Mizoguchi et al., 1996). An Activity has four components:

* Objective: refers to one or more Goal_Item through a "Links to" relationship. For example, the specific Objective of an Activity whose type is "Exploring" can be linked to a Goal_Item of type "Memory."

* Prerequisite: describes the conditions required to perform the Activity. For example, a "Basic Explorer" skill is needed before starting a "Tutoring" Activity.

* Task: represents the minimum LS working unit. In (Mizoguchi et al, 1996), there is an educational task taxonomy, which has been used in the definition of the Task entities (e.g., "Explain," "Identify," "Suggest," or "Visit").

* Result: specifies the possible values returned by the Task elements. For example, a "Visit" Task may return the number of visited items, the time spent in each item or the order followed to visit the items.

More specific and context-dependent information can be added to the Activity entities when they are linked to the certain Didactic Structure entities concerning a specific domain of knowledge.

The Management entity aggregates two types of elements:

* Conditions to describe the general aspects that control the scenario operation, such as the Interaction mode (for example; the kind of human-computer interface to interact with the application or the portrayal configuration to display the content items using specific media) or the Control parameters (e.g., whether the navigation through the contents is free or guided).

* Outcomes to represent the Activity outputs such as Performance levels and Action items. Since the values of both outputs are implicitly vague, they can be "granulated" using fuzzy categories (Zadeh, 1997) like {"very high," "high," "normal," "low," "very low"} for the Performance levels and {"completed actions," "semi-completed actions," "incomplete actions"} for Action items. These categories, expressed through linguistic labels and qualifiers are properly represented as fuzzy sets just as it is done in a large number of existing applications. In this way, fuzzy logic could be incorporated to the model to analyze the task outcomes and provide an adaptive reply to the user interaction.

There are also semantic relationships between Management and Activity entities, such as Assigns that configures the Control parameters for each Task and connects Performance levels to the Activity Results. For example, if the number of "Visited" items is higher than a certain value, then the Action assigned to the "Exploring" Activity can be set to "Completed." Again, fuzzy rules could be incorporated here to establish a relationship, for example, "If the number of visited items is high and each visit is long enough, then the Activity has been completed."

As a summary, Figure 5 shows the diagram (including values for the attributes of each entity) corresponding to an example of Learning Scenario called "Introduction" which could serve as an introductory lecture for several knowledge domains. The example has two final "Memory" Goal_Item ("Short-term" and "Long-term memory"), which are based on a "Domain_Independent" goal. There is a "Tutoring" Activity that "links to" a "Short-Memory" Objective and no Prerequisite entities are defined. Basic Tasks are used such as "Visiting" the content nodes or "Explaining" some aspect related to these nodes that requires a more elaborated information or an evaluation of the learner's understanding. Result elements describe the number of "visited" nodes. In the LS Management some Interaction features are defined such as a "free" navigation or a "multimedia" portrayal configuration. There are also specified Performance elements; one of them is connected to a Competency item, which is associated to a Learner profile (see description). This Learning Scenario can be used to access to knowledge organizations.

[FIGURE 5 OMITTED]

Didactic structure. The last of the three main entities of the model is the Didactic Structure (DS), which is defined as a set of educational resources and contents organized in a didactic way. It is composed by a Knowledge-StructureSet entity which, in turn, aggregates one or more Knowledge Structure (KS) entities and an InstructionalTaskSet entity which also aggregates one or more InstructionalTask (IT) entities (see the UML class diagram in Figure 6).

[FIGURE 6 OMITTED]

KS entities are based on the concept of Knowledge Structure proposed in (Merrill & ID2 Research Team, 1996) and they are used to organize the DS contents in units called Knowledge Objects (KO). There are several types of KS such as "List," "Part Taxonomy," or "Dependency" as it is described in (Merrill & ID2 Research Team) and each of them aggregates a number of KO entities. These entities are characterized by attributes as the identifier or the type ("Entity," "Process," or "Action" as defined in Merrill & ID2 Research Team), and with elements such as:

* Portrayal element to represent the media resources used to describe the KO item. For example, an "Image" used to describe a "Disk" device.

* Property specifies a KO property. For example, the "Disk" storage size that in turn can be presented using different portrayal elements (text or an image, for example).

* Keyword specifies concepts related to the KO item. For example, concepts such as "information storage" or "computer component," which are related to the "Disk" device.

IT entities are addressed to manage the access to the DS contents in a didactic way and they are based on the concept of "transaction shell" introduced in (Merrill & ID2 Research Team, 1996). With this purpose, they use Instructional Objects (IO) and Instructional Parameters (IP). IO entities are used to manage Merrill's knowledge objects (KO entities) from a learning point of view and IP values represent the conditions in which these KO entities are accessed, such as the abstraction level or the portrayal configuration. A didactic knowledge taxonomy proposed in (Leidig, 2001) is used to represent the IO entities.

IO and KO entities are related through "Links to" relationships. If the KO is an "Entity," then IOs such as "Definition," "Explanation" or "Example" can be used to describe or identify it. Similarly, "Action" KOs can be characterized using IOs such as "Exercise," "Question," "Analogy" or "Hint," and finally, "Process" KOs are associated to "Simulation," "Feed-back," or "Animation" elements.

The link between instructional and knowledge objects can be specified in a dynamic way. For example, a "Definition" element can be configured using several media resources (e.g., text or images) for the same knowledge item (e.g., a "Magnetic Disk" entity) and they can be dynamically selected at runtime depending on user preferences or needs.

IT entities are connected with KS entities using a "Links to" relationship. It represents the way IT entities access to the KS components. For example, a "Visit" IT can be used in several KS types: a "Part-Taxonomy" structure which permits a direct access to its KO components or a "Dependency" structure which configures a sequential access.

Figure 7 shows the UML object diagram that represents an example of DS oriented to the learning process about magnetic disks in a computer architecture course for which two KS entities: a "Part-Taxonomy" KS that describes the magnetic disk components; and a "Dependency" KS which sets up the sequence of knowledge items needed to understand how it is working the file physical storage on a disk. The "Part-Taxonomy" has a root element labeled "MgDisk" and some elements of the hierarchy (not all for readability purposes) are displayed, for example, those concerning the "Mechanical System" or the "Information System." They are represented by "Entity" KOs, which are linked to IO elements such as an "Example" or an "Explanation." There are two IT entities attached to this DS:

* "Visit": is based on accessing to the descriptive data about the KS components using different IO entities and instructional parameters. In Figure 7, an IO "Example" is used in order to show images of the "Disk Arm" component with a low level of depth (associated to the "Depth" Instructional Parameter value).

* "Explain": is a more complex IT that involves "Explanation" IOs, which can be attached to other IO entities such as "Activity" or "Question" items. These IO entities allow the user to interact with KO components in order to check if the presented content was understood. The returned IT values can be used to control the navigation through the KS components (e.g., the next step in a "Dependency" structure).

[FIGURE 7 OMITTED]

Relationships among learning scenarios and didactic structures. Previous sections were devoted to describe the entities that refer to the learning process (Learning Scenario entities) and the contents they access to (represented by Didactic Structure entities). UML diagrams have been used to represent them individually and from an abstract point of view. One of the main advantages of this separation is the possibility to define generic and domainindependent LS entities (Paquette et al., 2001) and to apply them to different educational contents, in a dynamic way. The proposed dynamic link mechanism is based on the semantic relationships between LS and DS entities, which are highlighted in the UML diagram displayed on Figure 8:

[FIGURE 8 OMITTED]

* "Links to": connects a LS Task with the implementation of an Instructional Task in a Didactic Structure context.

* "Defines": selects the Interaction values, which are used as parameters of the Instructional Task.

* "Assigns": selects the Instructional Object elements, which are used as Activity Result entities.

The specification of these semantic relationships is independent from the Learning Scenario and the Didactic Structure entities so that instructors can check the Task operation in different scenarios without modifying such entities.

Operational Model

The previous section figures out the static description of the model entities and their main relationships, both structural and semantic. The current section describes how these entities evolve and behave in a learning environment and how users interact with them.

The behavior specification provides the means to check the instructional properties in IA products without the cost of implementing them. For example, we can check how user objectives can be fulfilled from the goals defined in Learning Scenario entities or how Tasks make use of different Interaction parameters in their operation. The operational model is described through several abstraction levels that use different UML notations: Use Cases and Sequence diagrams. The higher abstract level is based on UML Use Cases to illustrate the processes that users can perform. Figure 9 shows a Use Case diagram for an IA.

[FIGURE 9 OMITTED]

As can be seen, there are several user types and each of them has been associated with a specific function. For example, a Learner user accesses learning processes, which are represented by LS entities. These entities can be defined by a Pedagogue in a generic way or by a Teacher to implement learning activities related to a more specific knowledge domain. An Author develops educational contents that a Teacher uses to create specific DS entities for organizing a given domain of knowledge. Meanwhile, a Tutor can supervise the interaction between Learner users and LS entities. Users can also interact with external entities such as ECP (Educational Content Providers such as an IMS-based repository [Burguillo et al., 2002]) and ESP (Educational Services Providers such as a third part Learning Management System (Agusti, Buendia, & Benlloch, 2002)).

Some of the functions performed by users are related to the LS management, such as "Design learning processes" (Pedagogue user), "Access learning processes" (Learner user) or "Control learning processes" (Tutor user), or the DS management such as "Organize didactic contents" (Teacher user). For example, Figure 10 shows the UML Use Case diagram assigned to the Learner user that displays cases such as:

* "Open session": the Learner starts a working session and he obtains the information about his identification and other personal data which are stored in User Profile entities or come from external Educational Service Providers. He/she can also state his objectives for the current session.

* "Search LS": the Learner can explore the available LS entities he is interested into, using searching criteria such as the User Profile preferences, the objectives stated in the current session or the outcomes to be obtained.

* "Select LS": the Learner decides the current LS from those found in the previous case or active LS started up in previous sessions.

* "Select Activity": in a given LS example, several Activity entities can be available and data analysis about Objective or Prerequisite items helps the Learner to select one among them.

* "Perform Activity": the Learner resumes or starts a new Activity by running its component Tasks. Prerequisite information is compared with Learner competencies to check if the Learner is ready to perform the selected Activity. He can also finish the Activity; then, its Results are recorded.

[FIGURE 10 OMITTED]

* "Close session": once the Learner finishes the current session, Activity results are checked to determine the obtained Outcomes and the Competency items to be registered in his User Profile.

Instructional application sequences. A second level of behavior description of the model is based on UML Sequence diagrams that specify entity interactions in a time sequence. They make possible a more detailed description of the use cases that display the IA working.

Figure 11 shows an example of Sequence diagram, which describes the "Search LS" case and more specifically the "By Objectives" sub-case (see Figure 10). It represents the time sequence of events, on which the Learner user is involved when is searching a Learning Scenario entity which goals match his personal objectives. The first event represents the selection of the searching criteria based on user learning objectives. Next, the Learner user accesses to his Objective list that is stored in the User Profile entity and selects one (or more) of them. The selected Objective item is matched up to the "User Goal Relationship" entity, which implements the relationship between user objectives and general learning goals. These goals are then compared with the LS GoalsStructure entity to check if some of them are fulfilled.

[FIGURE 10 OMITTED]

The last event returns those Learning Scenario entities that meet the user requirements. This selection is based on the automatic processing of the "User Goal Relationship" entity or supervised by a Tutor user (not included in the example diagram). The rules or conditions that control this processing are assigned by the Teacher or Pedagogue users and these rules are part of the "User Goal Relationship" definition. Sequence diagrams can be also applied in other contexts and they have been used to specify the behavior of some of the entities that belong to the current IA model. This specification helps to the definition of properties such as:

* "Goal Fulfillment": checks the instructional application ability to satisfy selected user objectives.

* "Task Results": checks the results returned by the instructional tasks that are linked to a specific DS entity.

A tool has been developed to visualize the specification of the model entities as well as to check their dynamic behavior. Figure 16 shows a tool screenshot that displays the result of the checking the "Goal Fulfillment" property. This operation allows the instructor to check the goal adaptation possibilities of the specified IA. In this case, the "Get Technical Skills" objective is matched against the "LSBasicCourse" LS. The lower area displays the goal trace in which LS goal items that match the user objectives are marked.

[FIGURE 16 OMITTED]

DISCUSSION

The framework introduced in the previous sections is related to the concept of Educational Modeling Languages (EML) that has been defined in (Cen/Isss, 2002) as: "a semantic information model and binding, describing the content and process within a 'unit of learning' from a pedagogical perspective in order to support reuse and interoperability." This section reviews some EML proposals and compares them with the model presented in this article.

Learning Material Markup Language (LMML)

LMML is based on a meta modeling architecture for knowledge management (Suss, Kammerl, & Freitag, 2000). LMML is addressed to describe the content of units of learning in arbitrary domains of knowledge. Within its information model, a learning unit is organized modularly and it consists of various modules, which in return may contain further modules. The basic units of information in this module-hierarchy are called "ContentModules" (e.g., definition or motivation) that are similar to the Instructional Object entities proposed in the current framework. They contain "Media Objects," for instance, pictures, animations or text that can be structured as lists or tables ("Structure Objects").

The LMML information model is focused on structuring the content of units of learning but these structures are bounded to hierarchical organizations. Learning activities are not explicitly represented in the LMML notation and external models (Weitl, Suss, Kammerl, & Freitag, 2002) described the access to the LMML units of learning.

PALO

PALO (Rodriguez, Verdejo, Mayorga, & Calero, 1999) defines a cognitive approach of an EML that describes courses structured in modules. Each module includes a declaration of the structure, the activities students and tutors undertake and the scheduling of activities and contents. The Palo information model is structured in the next levels: Content, Activity, Structure, Scheduling, and Management.

The Content level manages entities such as "Bd_object" that represents concepts or instructional objects (e.g. examples or definitions), or "Bd_relation" that represents didactic relationships between "Bd_objects" (e.g. a concept prerequisite). "Bd_objects" include any type of media content.

The Activity level uses the concept of "Task" to represent any kind of learning process and the Structure level organizes Content and Activity entities by means of elements such as "Module," "Part," or "Subpart." The Scheduling level assigns deadlines and prerequisites for Structure and Activity entities. The Management level refers to the databases that store the Content and Activity information and the metadata that describes the course document.

OUNL-EML

OUNL-EML (Koper, 2001) has been developed by the Open University of the Netherlands for use in e-learning and it has been recently adopted by IMS as the basis for the Learning Design project (IMS, 2002). OUNL-EML is based on a pedagogical meta-model, which uses the concept of "unit of study" as the smallest unit providing learning events. It aggregates entities such as:

* "Role" that represents the actors of the model (e.g., learner or staff)

* "Learning objectives" and "Prerequisites" that are required by the "unit."

* "Activity" that represents any kind of learning process.

* "Environment" that contains all the knowledge and other types of objects which are used by "Activities."

* "Method" that controls the "Activity" played by the "Role" entities.

The "unit of study" also embeds semantic relationships between previous entities, such as: a "Role" entity "performs" an "Activity," or this "is performed" in an "Environment" entity, or a "Method" entity is "designed towards" matching some "Learning objectives."

EML comparison. The comparison with the previous EML proposals is performed using the next criteria coming from the EML definition (Cen/Isss, 2002):

* Information model.

* Content representation.

* Process representation.

A first contribution of the current proposal is the introduction of an operational model that complements the semantic information model required in any EML proposal. Palo and OUNL-EML provide several use cases but their description is informal. In the current operational model, two formalization levels are defined, based on UML Use Cases and Sequences, and they have been implemented using XML notations.

The semantic information model is also divided into three main issues: one related to processes (Learning Scenarios); the second one that manages contents (Didactic Structures) and the third one, which is concerned to the user learning aspects (User Profile). Palo considers the specification of processes and contents, and only OUNL-EML addresses the full set of issues. Nevertheless, those proposals modeled such issues in a declarative way and the possibility of dynamic links between entities was neglected. In the current proposal, relationships between entities (e.g., Learning Scenario and Didactic Structure, or User Profile and Learning Scenario) are modeled explicitly and they can be managed at the operational model.

Regarding the content representation, the reviewed EML proposals agree in the separation between instructional-didactic (Instructional Object entities in our model) and knowledge (Knowledge Object entities) information. This feature promotes reuse but hides the relationships that educational ontologies introduce, such as those coming from the didactic knowledge taxonomy proposed in (Leidig, 2001). For example, the relationship between an "Example" and an "Explanation" item, and the use of different media to implement these items. The Palo proposal defines connections between instructional items but they are attached to specific contents. In the current proposal, the relationship between IO and KO entities is defined alongside these entities allowing for example, the media configuration of an "Explanation" item to be selected at runtime, depending on the user preferences or needs.

The process representation is restricted to the Palo and OUNL-EML proposals as it is described in (Cen/Isss, 2002). OUNL-EML provides a more flexible and powerful way to represent and organize learning processes, using "Activity-Structure" entities, but both proposals share a common feature: they link the process units ("Task" in Palo and "Activity" in OUNL-EML) to the contents they access, in a static way.

In the current proposal, the learning processes are represented by Task items that are grouped by Activity entities. They can be dynamically linked to the Instructional Task entities that implement them within a knowledge domain (Didactic Structure context). Therefore, the same contents can be accessed from a learning scenario whose requirements change at runtime. For example, the same "Visit" Task that accesses the components of a "Part-Taxonomy" Knowledge Structure would require a sequential or a direct access to these components, depending on the current Learner progress.

CONCLUSIONS

This article introduces a framework for specifying instructional applications beyond the rigid courseware structures underlying the typical web-based courses. The proposed framework is based on an Instructional Application model, which provides a gateway between instructional design concepts and computer-based notations.

This model considers two main matters: first, a semantic view that describes the model entities and their relationships. This view is organized into three main entities: Learning Scenarios that specify learning processes and their context; Didactic Structures, which organize the educational resources and contents and User Profiles to represent the user information concerning learning and teaching aspects. This separation eases the dynamic link between these entities, using elements to model their relationships and providing adaptiveness features. These model entities are also composed by other elements such as Activities, Tasks, Knowledge Structures, Knowledge Objects, Instructional Objects, Preferences, or Competencies, which can be reused in different applications.

The second perspective is used to address the model behavior and how users can interact with the model entities. Such an operational point of view allows instructors to detail procedures for managing the model entities and their relationships. In this way, instructors will be able to specify instructional issues such as the rules for analyzing user needs and objectives, the implementation of learning tasks, or the competency assignment from the task outcomes.

Another main contribution is the ability to check these instructional specifications without the cost of their implementation, by means of tools such as the one mentioned in the current work.

Further works consider issues such as:

* The development of authoring tools to allow users (e.g., Pedagogue, Tutor, or Teacher) the edition of model entities through a simple and visual interface since in the current status, the model is hard to be used.

* The outline of a design methodology to guide developers in the application of the presented model, with the aim of generating instructional products.

* The definition of a full set of checking criteria for the evaluation and comparison of the specification of the model entities and their behavior.

APPENDIX

ELEMENT                                EXAMPLE

Composition. Structural relationship            The same ActivitySet
that implies that an element is                 cannot appear in two
composed by other elements in such a            different Learning
way that a specific instance of a               Scenarios
component cannot take part in
different compositions.
Aggregation. Structural relationship            The same Activity can be
that implies that an element is                 reused in several
composed by other elements in such a            ActivitySet entities
way that a specific instance of an
aggregated element can take part in
different aggregations.
Generalization. Structural                      An Outcome is a general
relationship that implies that an               concept that can be
element is specialized into a number            specialized as a
of elements that inherit its                    Performance item or an
properties and behaviors.                       Action
Association. Semantic relationship              An Objective refers to a
among elements.                                 Goal Item
Multiplicity. Minimum and maximum               A Knowledge Object has
cardinality of a relationship. The              from 0 to n Properties
symbol * is used to represent any
number of occurrences.
Role. Name given to a specific                  An Instructional Object
relationship to increase its                    has a "Links to"
semantics.                                      relationship with
                                                another entity
Actor. Someone or something that is             A Learner user which
external to the system, but that is             uses LS entities for
going to interact with the system.              learning.
Use Case. Sequence of related                   A sequence of
transactions performed by an actor in           transactions related to
the system.                                     the session opening.
"Includes" Case. A case that includes           An "open session" case
other cases.                                    can be divided in
                                                several cases
"Generalized" Case. A case that can             There are several types
be instantiated in several case                 of "LS Search"
types.
Object Line. Sequence of temporal               The User Profile can
events that concern an element.                 manage read access
                                                events
Sending Message. Event that transmits           The User Profile
data or control information to                  notifies the request of
another element.                                Preference information
Returning Message. Event that                   The User Profile
receives data or control information            receives the Competency
from another element.                           information

References

Agusti, M., Buendia, F. Benlloch, J.V. (2002). XEDU, an implementation of an open learning management system. Proceedings of the Third International Conference on Information Technology Based Higher Education and Training July, 2002. Budapest, Hungary.

Booch, G., Jacobson, I., & Rumbaugh, J. (1998). The unified modeling language. Boston, MA: Addison-Wesley. 1998.

Brusilovsky, P. (1998). Adaptive educational systems on the World Wide Web: A review of available technologies. Proceedings of Workshop "WWW-Based Tutoring" at 4th International Conference on Intelligent Tutoring Systems. San Antonio, TX, August 16-19.

Buendia, F., & Diaz P. (2002). A framework for educational adaptive hypermedia. Proceedings of Adaptive Hypermedia and Adaptive Web-Based Systems. Second International Conference, May 2002, pp. 476-479. Malaga, Spain.

Burguillo, J.C., Buendia, F. Benlloch, J.V., Rodriguez, D., Gomez, J.M., & Vidal, J.J. (2002). Tools for creation and management of didactic resources in EIE. Vienna International Working Conference--eLearning and eCulture. VIEWDET December, 2002. Vienna, Austria.

Burns, H. L. & Capps, C. G. (1988). Foundations of intelligent tutoring systems: An introduction. Hillsdale, NJ: Lawrence Erlbaum.

Cen/Isss (2002, July). Learning Technologies Workshop. Educational modelling languages. European Committee for Standardization/International Society Standardization System. [Online]. Available: http://www.cenorm.be/isss/Workshop/lt/

IMS (2002). IMS Global Learning Consortium [Online]. Available: http://www.imsproject.org/

Kasowitz, A. (1998, August). Tools for automating instructional design. ERIC Clearinghouse on Information & Technology [Online]. Available: http://www.ericit.org/digests/EDO-IR-1998-01.shtml

Koper, R. (2001). Modeling units of study from a pedagogical perspective: The pedagogical meta-model behind EML. [Online]. Available: http://eml.ou.nl/introduction/docs/ped-metamodel.pdf

Landoni, M., & Diaz, P. (2003). Special issue on e-education: Design and evaluation for teaching and learning. Journal of Digital information, 3(4).

Langevin Learning Services (2002, November). Langevin Instructional DesignWare. [Online]. Available: http://207.107.10.214/lid.asp

Leidig, T. (2001, March). L3' towards an open learning environment. Journal of Educational Resources in Computing (JERIC), 1(1).

Lindner. R. (2001). Expertise and role identification for learning environments (ERILE), 2001. [Online]. Available: http://jtc1sc36.org/doc/36N0065.pdf

Mentergy (2002, November) Designer's edge. [Online]. Available: http://www.mentergy.com/products/authoring_design/designer/.

Merrill, M. D., & ID2 Research Team (1996). Instructional transaction theory: Instructional design based on knowledge objects. Educational Technology, 36(3), 30-37.

Mizoguchi, R., Sinitsa, K., & Ikeda, M. (1996). Task ontology design for intelligent educational/training systems. ITS'96 Workshop on Architectures and Methods for Designing Cost-Effective and Reusable ITSs. Montreal, Canada.

Nervig, N. (1990). Instructional systems development: A reconstructed ISD model. Educational Technology, November 1990, 40-46.

Paquette, G., Rosca, I., De la Teja, I., Leonard M., & Lundgren-Cayrol, K.(2001). Web-based support for the instructional engineering of elearning systems. WebNet'01 Conference. Orlando, FL, USA.

Rodriguez-Artacho, M., Verdejo, M.F., Mayorga, J.I., & Calero M.Y. (1999). Using a high-level language to describe and create web-based learning scenarios. Frontiers In Education Conference FIE '99 IEEE Computer Society, San Juan, Puerto Rico.

Suss, C., Kammerl, R., & Freitag, B.A (2000). Teachware management framework for multiple teaching strategies. Proceedings of Educational Multimedia and Hypermedia. EDMEDIA'00, Montreal, Quebec, Canada.

Weitl, F., Suss,C., Kammerl, R., & Freitag, B.A (2002). Presenting complex e-learning content on the web: A didactical reference model. Proceedings of E-Learn 2002 World Conference on E-Learning in Corporate, Government, Healthcare, & Higher Education, 2002.

Zadeh, L.A. (1997). Towards a theory of fuzzy information granulation and its centrality in human reasoning and fuzzy logic. Fuzzy Sets and Systems, 90(2), 111-127.

FELIX BUENDIA-GARCIA

Universidad Politecnica de Valencia

Spain

[e-mail]

PALOMA DIAZ

Universidad Carlos III de Madrid 

Spain

[e-mail]