Application of Semantic Web Concepts to Intelligent Tutoring Systems Editor(s): Name Surname, University, Country Solicited review(s): Name Surname, University, Country Open review(s): Name Surname, University, Country

Aparna Lalingkar,a, and Chandrashekar Ramnathan,a a International Institute of Information Technology, Bangalore, India E-mail: {aparna.l, rc}@iiitb.ac.in

Abstract. The application of concepts of Semantic Web to Intelligent Tutoring directs towards Adaptive Educational Hyper-media Systems (AEHS). AEHS contains a document space (i.e. domain ontology), a user model (i.e. a student model ontolo-gy), a set of observations (i.e. the logic through which students’ interaction with the system can be observed), and the adapta-tion component (i.e. suggested hyperlinks or even feedback to the student about what the student knows and still needs to learn). Reusability and interoperability offered through resource description framework (RDF) and ontology increase the scalability of the potential use of Intelligent Tutoring Systems (ITS). There are several research papers which discuss the bene-fits of the use of Semantic Web concepts for user modeling in Intelligent Tutoring Systems. There is a lot of research going on in the AI-ED community regarding enhancing intelligence in the ITS. There are some ITS and authoring systems in which Semantic Web concepts are used. In this paper the basic literature on ITS, AEHS and AI-ED is reviewed. Also, the basic con-cepts of Semantic Web are reviewed and its connection with research in all the above fields is examined. It is found that the combined use of research insights received from all the three fields, together with Semantic Web technology and AI tech-niques, would enhance intelligence for tutoring systems on the Web. This provides a basis for further research in the field of Intelligent Tutoring aimed at enhancing the intelligence of the systems.

Keywords: Intelligent Tutoring Systems, Adaptive Educational Hypermedia Systems, Ontology, Personalization, Semantic Web, AI-ED

1. Introduction

Intelligent Tutoring Systems (ITS) offer computer-based personalized instructions by imitating a human tutor, and provide a highly efficient learning envi-ronment [1]. Traditionally, ITS have four components [2, 3], namely: Tutoring Module (Pedagogical Mod-ule), Expert Module (Domain Knowledge), Student Module (Student Model) and Interface Module (Prob-lem Solving Environment or User Interface), through which the student interacts with the system. While analyzing the state of the art of authoring ITS, [4] has mentioned Intelligent/Adaptive Hypermedia Systems as one of the authoring ITS. For more than fifteen years, many researchers [5-10] have studied Adaptive Hypermedia Systems with special applications to web-based education, which come from the combina-tion of research in hypermedia systems and user-

modeling. Adaptive Educational Hypermedia Sys-tems (AEHS) contain a document space, a user model, a set of observations, and an adaptation component. Adaptation in AEHS, i.e. use of user models and adaptive component, added “dynamic” value in tradi-tional “static” hypermedia systems by providing dif-ferent educational content to different users based on their requirements.

Recent developments in Web Technologies and ef-forts to make them more intelligent led to the devel-opment of Semantic Web Technology. “The Seman-tic Web is a vision, the idea of having data on the Web defined and linked in such a way that it can be used by machines not just for display purposes, but for automation, integration and reuse of data across various applications”1

1

. The Semantic Web is for im-

http://www.w3.org/2001/sw/Activity/

plementing reliable, large-scale interoperation of Web services and making such services computer-interpretable to create a Web of machine understand-able and interoperable services that intelligent agents can discover, compose and execute automatically [11]. The Semantic Web offers scalability and ex-pands the services of the existing web by enabling software agents to automate procedures currently performed manually and by introducing new applica-tions that are infeasible today [12].

Application of Semantic Web concepts to Intelli-gent Tutoring directs towards AEHS because of simi-larity between the concept of student modeling and user modeling. Reusability and interoperability of-fered through Resource Description Framework (RDF) and ontology increase the scalability of the potential use of ITS [13]. There are several research papers [14-19] which discuss the use of Semantic Web concepts for user modeling in Intelligent Tutor-ing Systems. There are some AEHS, such as OWL-OLM [20], OntoAIMS [21] and ActiveMath [22], in which Semantic Web concepts are used.

In this paper, the basic literature of ITS and AEHS and the basic concepts of Semantic Web are first re-viewed. This review is followed by discussion of the connection between AEHS and ITS when concepts of Semantic Web are used for Intelligent Tutoring. Fi-nally, some AEHS, which use basic concepts of Se-mantic Web, are reviewed, discussed and classified to get an overview for further research.

2. Intelligent Tutoring Systems (ITS)

Intelligent tutoring systems are the application of Artificial Intelligence in the field of education. In the literature, tutoring generally means intelligent tutor-ing. According to [23, 24], ITS are computer-based instructional systems in which instructional content and teaching strategies are modeled in such a way as to specify what to teach and how to teach, respective-ly. Four components in the structure of an ideal ITS are described in Figure 1. In an ITS, the student is engaged with the interface module in order to get the actual tutoring. In the tutoring module, models of pedagogical strategies are included. The domain knowledge module is nothing but the expertise mod-ule which is used as a knowledge base for teaching. The student module (i.e. the student model) is infor-mation about the student’s interaction with the sys-tem: what the student has learned, which exercises the student has solved, etc. Basically, the student module is a diagnostic module which helps to diag-

nose students’ misconceptions and provide feedback [25]. The expert module and the student module are discussed in detail; this will help us to draw parallels between ITS and AEHS with respect to use of Se-mantic Web concepts.

Fig. 1: Structure of an intelligent tutoring system [2, 3]

2.1. Expert Module

For designing an intelligent system, one needs to have an abundance of domain knowledge. Domain knowledge can be encoded in three ways [26]. Firstly, there is the black box model, where only input-output information is available but no information is availa-ble regarding how the input information is processed to get the output. Secondly, there is the glass box model, which uses knowledge engineering tech-niques: i.e., it builds the knowledge model by con-sulting an expert. Finally, there are cognitive models, i.e. the way human beings use knowledge by coding different knowledge types. Cognitive modeling is used in Cognitive Tutors [27] (which are based on the Adaptive Control of thought Rational (ACT-R) theo-ry developed by [28]), and is one of the examples of the third way of knowledge encoding. ACT-R theory consists of two parts: declarative knowledge and pro-cedural knowledge. Declarative knowledge is nothing but factual knowledge stored in the form of chunks. Normally, chunking means dividing into small groups, but from a machine process ability perspec-tive, chunks are group of prepositional structures connected with each other. Procedural knowledge is the knowledge displayed in our behavior and uncon-sciously implemented by production rules. Produc-tion rules are formalisms for representing reasoning in mathematics [29], and are the rules required for performing different cognitive tasks. One of the ad-vantages of the rule-based approach is that it can be used for representing the student’s knowledge.

2.2. Student Module

Besides assessment of the student’s knowledge, the human tutor uses the student’s body language, facial expressions, and voice tone to recognize the student’s emotional state and then alter the teaching strategy as per the student’s need. This is the major idea behind adaptive intelligent tutoring, which is discussed in more detail in the section on AEHS. In the ITS, this information about the student’s compe-tencies, learning achievements, affective characteris-tics like emotional states, opinions, and learning is gathered from the student’s interaction with the sys-tem and stored in the student model [30].

Following are the major reasons described in [31] for why the student model is used in ITS. Firstly, for learning a structured curriculum through an ITS, if the student wants to move on to the next topic s/he needs to master the topic which is the prerequisite of the next topic. Student modeling is used to determine which topics the student has mastered. Secondly, to give spontaneous advice at appropriate places, the system must know the student’s previous knowledge, which can be stored in the student model. Thirdly, based on the student’s capabilities known through the student model, some ITS generate problems suitable for that particular student. Finally, the student model is used for giving adaptive feedback based on what the student already understands, a record of which is stored in the student model. If students are provided customized feedback and are taught with instructions tailored to their individual learning styles, their inter-est in learning may be increased [30].

Classification of student models in a three dimen-sional space has been made by [31], the three dimen-sions being: bandwidth, target knowledge type and differences between student and expert. The band-width dimension is related to the quality of input. This bandwidth (which is nothing but the time the student takes to solve a problem or exercise) can be used for the diagnosis of the state of a student. There are three types of states: mental states, intermediate states, and final states, which are mentioned in [31] under the bandwidth dimension. The other two di-mensions are related to structural properties. Three target knowledge types are mentioned: flat procedur-al, hierarchical procedural, and declarative. The complexity of the interpretation of each type of knowledge affects the diagnosis, and hence the stu-dent modeling. According to [31], flat procedural knowledge makes student modeling easier, hierar-

chical procedural knowledge makes it more complex, and the declarative makes it most difficult.

There are two types of differences between student and expert: missing conceptions and misconceptions [30, 31]. Missing conceptions are the ones which the expert has but the student does not have, and miscon-ceptions are the ones which the student has but the expert does not have. For determining missing con-ceptions, the student model is considered as a subset of the expert model, which is also called overlay model. For determining misconceptions, a separate list of possible misconceptions is used, which is called a bug library. This bug library can be created by using literature, by observing students, or by up-dating it in real time during the student’s interaction with the ITS. It is better to create a bug library by diagnosing the student’s misconceptions when the student is interacting with the system than to use a pre-defined one. To diagnose the misconceptions, one can predefine the library of predicates and the actions drawn from predicates. This predefined library of predicates and actions is called bug part library. This has stronger diagnostic capacity than the overlay model. Table 1 is borrowed from [32, p. 56] to give an idea about the variety of knowledge represented in student models.

Table 1

Variety of Knowledge Represented in Student Model [32]

2.3. Techniques for building Student Models

There are Cognitive Science techniques and Artifi-cial Intelligence techniques for building student mod-els. Model-tracing and constraint-based are Cognitive Science techniques, whereas formal logic, expert sys-tems, plan recognition or machine learning and Bayesian belief networks are AI techniques [30].

Model-tracing Technique: Cognitive Tutors [27] and the Andes Physics Tutoring System [33] are two of the systems which use the model-tracing approach to trace the student’s interaction with the tutors by using expert knowledge, thus determining whether students are heading towards an acceptable solution path or not [34]. The Model-Tracing Technique uses procedural rules of the following form: IF c1, c2 are true, THEN take the action a. Along with correct production rules, Cognitive Tutors have buggy pro-duction rules for tracking the student’s wrong ap-proaches. One of the limitations of model-tracing is that it does not talk about the student’s misconcep-tions and only tells whether the student is going in the right direction or wrong direction; this is because of the predefined nature of bugs. Another limitation is that if a student makes a mistake which is not in the predefined model, then it cannot be captured by the Model-Tracing Technique.

Constraint-Based Technique: Limitations of mod-el-tracing are due to the assumption that all mistakes by students can be captured and modeled for machine understanding. The constraint-based technique as-sumes exactly the opposite. The belief that students learn from their errors [35] is the basis for constraint-based techniques. SQL-tutor [36], WETAS [37], and KERMIT [38] are systems in which constraint-based techniques are used. According to [35], we make mis-takes because of lack of internalization of declarative knowledge into our procedural knowledge, and by practicing the task we modify our procedural knowledge and learn [39]. In constraint-based model-ing, the system is not interested in what the student is doing, but in the problem-state that the student is in. As long as the current state of the student is not iden-tified as wrong, the system does not interrupt the pro-cess and the student can proceed freely. The domain model contains a set of rules in the following form: “If <relevance condition> is true, then <satisfaction condition> had better also be true, otherwise some-thing has gone wrong” [39, p.315]. The relevance condition is a set of problem states for which the con-straint is relevant, and the satisfaction condition is the subset of states in which the constraint is satisfied. In

case of constraint violation, the system is expected to provide remedial action. The constraint-based tech-nique is based on the assumption that the student’s learning cannot be completely recorded and only er-rors can be recognized by a system [30].

Machine Learning Techniques: Student modeling techniques from Artificial Intelligence are based on reasoning about the knowledge available about the student, and not on human learning. Machine learn-ing techniques such as Bayesian belief networks, re-inforcement learning, supervised learning, hidden Markov models, fuzzy logic, and decision theory are used for student modeling [30]. Most of the tech-niques above use probability theory to predict student behavior and make tutoring decisions. Bayesian be-lief networks facilitate the combining of prior knowledge or belief with new data and offer a guide-line to keep new information in the proper position. In ITS, Bayesian belief networks are used for the classification of students’ knowledge, and for the prediction of their behavior and the steps where they will need help. There are two types of student models that can be generated by using Bayesian belief net-works: expert-centric and data-centric [40]. In the expert-centric model, networks and conditional prob-abilities are specified either directly or indirectly by experts; whereas in the data-centric, structure and conditional probabilities are learned mainly from the data collected from real-world evaluations.

Expert Systems Technique: In the field of AI, ex-pert systems (ES) are computer-based systems (pro-duction systems) that imitate the decision-making ability of a human expert [41]. They are designed for problem solving by using human-like knowledge expertise. Expert systems are composed of at least a knowledge base, an inference engine, and a user in-terface [42]. In ESs, production rules are used for diagnosing the problem and then suggesting solutions, which makes them a part of problem solving research. Expert systems have been researched and used exten-sively in medical diagnostic systems [43], intelligent patient monitoring systems, and management [44]. One of the major research links between expert sys-tems, knowledge modeling and problem solving is the work done by [44-47]. The importance of the role of ontology (see section 3.4) for knowledge modeling is highlighted by [48, 49]. Opal, a knowledge acquisi-tion system for oncologists and a tool for building Oncocin knowledge bases, was developed by [50]. Oncocin was an advice system for protocol-based cancer therapy developed by [51]. General develop-ment of Opal led to the development of Protégé-I, which succeeded in decreasing the difficulties of

knowledge acquisition for the purpose of medical advice [52]. Further, Protégé II was developed with the idea of reusable problem solving methods, and followed research by [47, 48] for the development of a tool for generating a hierarchy of problem solving methods and inference rules as separate components. Protégé has been developed into a tool named Proté-gé 4.0.22

2.4. Tutoring/Pedagogical Module

, which is also used for editing and creating ontology with a class-subclass hierarchy and a hierar-chy of properties and instances. AI techniques pro-vided the basis for the use of concepts of Semantic Web technology in ITS. How this is connected with Adaptive Hypermedia Systems, with particular refer-ence to Education, is discussed in section 4.

In the field of education, the term tutoring can be understood as an individualized instruction (for one, two, or three students at a time) in which continuous interaction takes place between the tutor and stu-dent(s). For tutoring, the tutor needs to adapt the tu-toring methods as per the requirement of the student and based on the interaction which the tutor has with the student. For this adaptation, the tutor needs to assess the student’s style of learning, diagnosing any errors in it, and then improve it by providing remedial feedback. Current research reveals that development in tutoring happens as a result of computing the stu-dent’s cognitive, meta-cognitive and affective knowledge [53].

In designing any educational application, the satis-faction of the four wives (what, when, where, who) and one husband (how) is necessary. In the case of ITS, the when question does not arise because it is a computerized system, and so the student can learn at any time. Four questions that one must think about before designing a tutoring module are discussed by [53, 54]: Who is the learner? (relating to the learner’s cultural and academic background, etc.); What is going to be taught during the tutoring? (relating to subject matter, etc.); How will the tutoring be done? (relating to teaching and the assessment strategies to be used during the tutoring, etc.); and Where or in which environment is the tutoring going to take place? (relating to whether it will be in the home or the school or college classroom, whether it is mobile learning or web-based learning, etc.). For exploring the answers to the questions which are mentioned above, one needs to model the tutoring according to

2 http://protege.stanford.edu/

the knowledge gathered from the student’s model. This is where we find the major connection between ITS and AEHS.

In order to design the tutoring module in such a way that conceptual understanding and self-regulated learning are included, one needs to consider pedagog-ical philosophy, such as constructivist learning, col-laborative learning, Bloom’s theory of mastery learn-ing, and Anderson’s ACT-R theory. Decisions about pedagogical theories and their effective use should be made after consulting pedagogical experts (i.e. teach-ers). Some researchers [55, 56] in the AI-ED (Artifi-cial Intelligence and Education) community have surveyed and analyzed the ITS. It was found that pedagogical experts were not involved in the author-ing of ITS as the authoring tools were not user-friendly. A huge gap was found between pedagogical planning and domain knowledge. The domain-specific nature of existing ITS was not sufficiently flexible that it could be reused. Even though ITS are an application of AI for education, the research community of ITS remained unfamiliar with the AI research community; this is why ITS researchers had never used developments in AI for improving intelli-gence in the ITS. Several researchers [55-59] have recommended and used Semantic Web concepts for authoring ITS (this is discussed in detail in section 6).

2.5. Interface Module

The essence of a good interface is that it should provide instruction and feedback in a clear and ex-plicit manner. The interface module must consider inputs from research on Human-Computer Interaction (HCI). Presentation of knowledge, reflection of in-structional objectives, use of more graphics, effective use of natural language (such as dialog), appropriate choice of display size, use of advanced computer techniques such as touch screen and speech recogni-tion, are some of the techniques discussed in [60, 53]. Presentation of knowledge through the interface plays an important role in the conceptual understanding of learners. While discussing the role of HCI in ITS, [60] proposed the breaking up of instruction as per the cognitive model, the use of dialog mode by the breaking up of instructions, the effective integration of graphics and natural language, and the presentation of semantic information in the domain for developing deeper understanding in learners. Discussing ad-vanced applications for ITS, [61] described experi-ments conducted by the MIT media Lab aimed at determining the learner’s affective domain during

interaction with the ITS; these experiments involved the use of various sensors, such as the mental state camera (for determining mental status), the skin con-ductance bracelet (for determining emotional status), the pressure-sensitive mouse, and the pressure-sensitive chair.

The basic concepts of Semantic Web technology are discussed in the following section.

3. Semantic Web Technology

Research into the application of AI concepts in Web development and efforts to develop the intelli-gent Web have resulted in the introduction of Seman-tic Web technology. The World Wide Web Consorti-um (W3C)3 is one of the leaders in Semantic Web Technology, particularly by the creation of standards. W3C has developed a Semantic Web layer cake (shown in Figure 2), which provides a basis for the development of Semantic Web technology.

Fig. 2: Semantic Web Layer Cake [55]

The Semantic Web can be built by exploiting the capacity of each block of the Semantic Web layer cake. Current research in the Semantic Web includes Linked Data [62], the practice of publishing and con-necting structured data on the Web that can be used for developing applications of Semantic Web in vari-ous fields. In the Semantic Web layer cake, the se-mantics increase from the lower to the upper level. Some of the sub-sections (such as 3.6, 3.7 and 3.8) are largely taken from the paper [63] in which these concepts had already been discussed by the authors of this paper.

3 http://www.w3.org/2005/Talks/1122-orf-sw/sw-stack-2005.png

3.1. URI and Unicode

The base layer consists of URI, Unique Resource Identification, which uniquely identifies a resource found anywhere on the Web. “Unicode is a compu-ting industry standard for the consistent encoding, representation and handling of text expressed in most of the world's writing systems” 4

3.2. XML

. In the Semantic Web, the Unicode is used to represent and connect documents written in different languages.

XML is Extensible Markup Language, which pro-vides a structured way to present data which can be read by machine. Markup Language is a modern sys-tem of annotating the text in a way that is syntactical-ly distinguishable from the actual text 5 . Markup means the sequence of characters, letters or other symbols that a document creator inserts at certain places in a text or word processing file to indicate how the file should look when it is printed or dis-played or to describe the document's logical structure. The markup indicators are often called "tags." For instance, in HTML (HyperText Markup Language, which is used for creating web-pages), to make a text for a page heading there are several tags, such as <h1> or <h2> or <h3>, which represent different levels of headings. XML Schema is the framework provided for the XML representation of any data. XML namespaces offer a way to globally use terms from more resources, as the Semantic Web needs to connect data from different resources. XML Query is a query language for extracting and manipulating the data in an XML document; it is also called XQuery6. Signature is a digital signature for authentication of a document which is written in the XML format7. En-cryption means basically the same thing that it means in cryptography, referring to the process of hiding private or secret information from others and making it accessible only if some special key is used8

4

. The encryption here is in the XML format, and is espe-cially developed as a standard for safety.

http://en.wikipedia.org/wiki/Unicode 5 http://en.wikipedia.org/wiki/Markup_Language 6 http://en.wikipedia.org/wiki/XML_query_language 7 http://www.w3.org/TR/2002/REC-xmldsig-core-20020212/ 8 http://www.w3.org/TR/2002/REC-xmlenc-core-20021210/

3.3. RDF

The next layers are RDF and RDF Schema. The Resource Description Framework (RDF) is a lan-guage for representing information about resources on the World Wide Web9. RDF contains a triple of subject, predicate and object. The RDF model can be graphically represented as shown in Figure 3.

Fig. 3: The RDF model

In the English statement Swati ate Mangoes, Swati is the subject, ate is the predicate, and Mangoes is the object. This statement can be written in the form of an RDF statement as shown in Figure 4.

Fig. 4: Example of simple RDF statement

RDF Schema is a framework that provides a basic structure to RDF. The RDF schema structure uses a special vocabulary of terms like classes, properties, container classes and properties, and RDF collections. In RDFS framework, terms such as Resource, Class, Literal, Datatype, XML Literal, and Property are used for generalized categories named classes. Terms such as range, domain, type, subClassOf, subPropertyOf, label, and comment are used for de-fining predicate categories or categories of properties. Terms such as Container, Bag, Sequence, and Mem-ber are used as container classes and properties. Terms such as list, rest, and nil are used for defining the RDF collections. These terms are generally used as follows, rdfs: Resources; rdfs: Class; rdfs: Datatype; rdfs:domain; rdfs:type; rdfs: subClassOf; rdfs: Bag; rdfs: Container; rdfs: Seq; rdfs: list; rdfs: nil. This means that these terms are used with refer-ence to their definition in the RDF schema. A collec-tion of such RDF triples can be used to bring about interoperability between applications that exchange machinereadable information on the Web. RDF is used for annotation of resources on the Web, which is

9 http://www.w3.org/TR/2004/REC-rdf-primer-20040210/

also called RDFa10

http://www.hogward.org/index.html

. The example shown in Figure 5 demonstrates the RDF/XML format as it is used with interconnections. There is a website

, which is created by a creator of value exstaff:65231. The same web-site has creation date of value April 1, 1860, and it is in the language with value en. This can be written in RDF/XML format as shown in Figure 5. In the above example, the terms xmnls and dc respectively stand for xml namespace (discussed in section 3.2) and Dublin core metadata (discussed in section 3.6).

Fig. 5: Example of RDF/XML format

3.4. Ontology

Generally the term Ontology means a philosophi-cal study of the nature of being, existence or reality11. In computer science or information science, ontology is a formal representation of a set of concepts within a domain and the relationships between those con-cepts12

For instance, in biology ontology, where facts in biology can be represented by using connections be-tween the concepts, Thing can be a super class, Liv-ing Things and Non-living Things can be two sub-

. Ontology is a specification of a conceptual-ization [64]. Conceptualization means a structured interpretation of a part of the world in which people think and communicate. Ontology is explained by [49] in two ways. Firstly, ontology is a representation vocabulary specialized to some domain, in which the terms in the vocabulary are intended to capture the conceptualizations of the domain. Thus, translation of an ontology from one language to another, say from German to English, does not change it conceptually. Secondly, ontology also refers to a body of knowledge describing some domain using a represen-tation vocabulary, where some facts about domain knowledge are represented.

10 http://en.wikipedia.org/wiki/RDFa 11 http://en.wikipedia.org/wiki/Ontology 12http://en.wikipedia.org/wiki/Ontology_%28information_scien

ce%29

classes of the super class Thing, and the class of Liv-ing Things can be further divided into two sub-classes, Vertebrates and Non-vertebrates. In other words, ontologies are essentially content theories representing the content in the form of classes and sub-classes of objects and the relations among them. Ontology contains hierarchies and tree structures. The spectrum for increasing semantics of ontology mentioned in [65] and described by [66] is shown in Figure 6. An ontology is defined by [67] as a set of knowledge terms including the vocabulary, the se-mantic relations among the terms, and some simple rules of inference and logic for some particular topic (this is quite similar to what is explained by [49]).

Fig. 6: The Ontology Spectrum [66]

There has been much debate about the best definition of ontology. The definition given by [65] is widely accepted, but one criticism of it is the general nature of the term specification, as there are no agreed-upon borders [68, 69]. Gruber’s definition of ontology is modified by [69] in the following words: “ontology is a formal specification of a shared conceptualization”. The inclusion of the two words ‘shared’ and ‘formal’ in the modified definition emphasize, respectively, the agreed upon conceptualization of the ontology that can be specified by a group, and the machine-readability of the ontology. Usually, ontologies can also be represented in the form of a set of RDF triples. They are used for information sharing and collabora-tion.

3.5. Logic and Reasoning

Logical reasoning is used to check the consistency of the data and infer conclusions which are not ex-

plicit in the ontology, and proof actually traces the steps of logical reasoning [70]. Trust stands for pro-vision of authentication and trustworthiness of the statements derived from the data. There are various languages used for ontology, such as Resource De-scription Framework Schema (RDFS), Web Ontology Language (OWL), Ontology Inference Layer (OIL), and DARPA Agent Markup Language (DAML). OWL is a W3C project for standardizing a more ca-pable ontology framework than RDFS [70]. OWL comes in three sublanguages named OWL Lite, OWL DL and OWL Full. OWL DL is based on Description Logic and OWL Lite is OWL DL with more re-strictions. OWL Full is a complete language with both OWL-Lite and OWL-DL. We can use Protégé to create ontology. After creation of the ontology, one can create rules by using the Semantic Web Rule Language (SWRL) 13

Logic is the third last building block of the Seman-tic Web layer (Fig 2). It is used for applying and evaluating the rules defined. It can help in inferring the facts that are not directly given in the ontology, and can interpret information which is not explicit. Logic can also detect contradictory claims and state-ments in the ontology [70]. The ontology, along with logical rules, forms a knowledge representation. Hence, it can be said that another role of logic is to represent and infer knowledge. There is a tool named RacerPro

, and by combining OWL and Rule Markup Language as defined by W3C standards.

14 (renamed ABox and Concept Expression Reasoner Professional), which can be plugged-in into the protégé; with the help of defined SWRL rules one can perform reasoning over the ontology. A-Box and T-Box are two notions that are frequently used in the world of ontology. T-Box stands for all the terms which are used in the ontology, which include clas-ses-subclasses, predicates, and the logical statements formed with them. A-Box includes all the instances of the classes and subclasses and the rules associated with the instances. RacerPro uses logic for reasoning the ontology through A-Box and T-Box. The ontolo-gy can be reasoned out without tools like RacerPro. As mentioned above, ontologies can be represented in RDF triple format. There is a query language named SPARQL 15

Apart from the basics of Semantic Web technology such as XML-RDF & Ontology, there are two other

(SPARQL Protocol And RDF Query Language) which can be used to query ontology in RDF format.

13 http://www.w3.org/Submission/SWRL/ 14http://www.racer-systems.com/products/racerpro/index.phtml 15 http://www.w3.org/TR/rdf-sparql-query/

important notions, metadata and learning objects, which are connected with Semantic Web and are used in AEHSs extensively. For completeness of the dis-cussion of the application of Semantic Web concepts in ITS, these two notions are discussed briefly in the next sub-sections.

3.6. Metadata

Metadata is data about data. In computer science, metadata is generally taken to mean information about a set of data in a particular representation, which typically means “information about objects” [71]. There are some metadata standards, such as Dublin Core16 and IEEE Learning Object Metadata (IEELOM)17, that are used all over the world. Dublin Core Metadata Element Set (usually referred to as just Dublin Core) is a metadata standard used by li-brarians to broaden the scope of book searches, and to facilitate search and information retrieval in cross-domain information resource description [72]. DC Metadata contains mainly bibliographic information, such as Title, Name of Creator, Subject, Description, Publisher, Contributor, Date, Type, Language, Source, etc. IEEE LOM standard is a principal metadata standard approved by IEEE, which contains seventy-six elements grouped in nine categories [73]. These have the nature of attribute and value. Sometimes metadata can also be seen as a fragment of ontology. Sharable Content Object Reference Model (SCORM)18

3.7. Learning Objects

is a collection of standards and specifica-tions for web-based learning that facilitates sequential content presentation and is commonly supported by learning management systems.

Learning can take place anytime and anywhere, whether in traditional classrooms, outside the class-room, or while working through educational videos or online courses. Learning benefits significantly from the availability of what the IEEE Learning Technology Standards Committee defines as Learn-ing Objects (LOs): that is, any entities, digital or non-digital, which can be used, reused, or referenced dur-ing technology-supported learning [73]. LOs stored anywhere, locally or globally, can be made available at any time and at any place to learners, provided that

16 http://en.wikipedia.org/wiki/Dublin_Core 17 http://en.wikipedia.org/wiki/IEEE_LOM 18 http://en.wikipedia.org/wiki/SCORM

they have network connectivity. LOs can be used in any learning environment and across disciplines, and facilitate the creation and use of online educational content [74]. LOs are flexible, portable, and adapta-ble, and can be used in multiple learning environ-ments and across disciplines. Learning objects are also reusable and interoperable [75, 76]. They may take a number of forms, such as a text document, a video/audio file, an image/animation, a simulation, or a virtual lab experiment. LOs can also be in the form of activities, games, or tests.

There are various Learning Object Repositories (LORs) available on the Web, such as Campus Alber-ta Repository of Learning Objects (CAREO) 19 and MERLOT 20. Learners can use the content of these LORs for updating and enhancing their knowledge. LORs store both LOs and their metadata [77]. The metadata maintains links to the associated LO if it is stored elsewhere on the Web [78]. Ontologies and metadata are both sharable. The IEEE Learning Technology Standards Committee has defined Learn-ing Objects Metadata (LOM) standards [73]. A num-ber of LORs, such as ARIADNE21, SMETE22, Learn-ing Matrix 23, MERLOT, CAREO, Learn-Alberta 24, and Lydia Inc.25

3.8. Topic Maps

, contain IEEE LOM profiles.

Topic Maps are covered by a number of papers [63, 79-85]. Generally, Topic Map technology is not ac-cepted as a part of Semantic Web technology by the Semantic Web community because it lacks the pow-ers of first order predicate logic for inference that are used in Semantic Web. However, Topic Maps do use XML-based schema, RDF for machine process abil-ity, the notion of ontology-instances, and some logi-cal inference by using Tolog query: these give it many of the capabilities of Semantic Web. Also, they are widely used by researchers [86, 87, 88] for web-based education. Hence, in this review, Topic Maps are also considered in the discussion of Semantic Web.

The basic concept model of Topic Map is TAO, where T stands for Topics, A for Associations, and O

19 http://en.eureka.ntic.org/display_lo.php?format=HTML&lom

_id=7742 20 http://www.merlot.org/merlot/index.htm 21 http://www.ariadne-eu.org 22 http://www.smete.org/ 23 http://thelearningmatrix.enc.org 24 http://www.learnalberta.ca 25 http://www.lydialearn.com/

for Occurrences [80]. An interesting facility provided by Topic Maps is the capacity to group Topics under Topic Types, Associations under Association Types and Occurrences under Occurrence Types. Also, there are Role Types which are used in the Topic Map to represent the roles played by different topics in the association. The use of a subject identifier for each topic in the form of a URI provides it with a namespace, i.e. a context. The facility of URI gives Topic Maps the power of merging and scalability. Various types (such as association types and role types) give Topic Maps the capacity to filter infor-mation. This is all explained in detail in [63, 79]. Topic Maps (TM) have general ontology and specific instances. Topic Maps were initially developed with a view to using them in the making of an encyclopedia. The advanced study and applications of TM in vari-ous fields revealed their usefulness for educational purposes. The standard used for creating Topic Maps is the XML-based Topic Map standard called XTM.

In the next section, basic concepts of AEHSs and some systems are discussed.

4. Adaptive Educational Hypermedia Systems (AEHSs)

Adaptive Hypermedia Systems are systems which provide user-driven services. “Adaptive Hyper Media Systems means all hypertext and hypermedia systems which reflect some features of the user in the user model and apply this model to adapt various visible aspects of the system to the user” [89]. AEHS have the capability of bringing about a paradigm shift of e-learning in two of its aspects: by the elimination of barriers of time, distance, and culture; and by the adaptation of the learner’s experiences [90]. Users with different goals, knowledge, background, experi-ence and preferences get service according to their personal needs. Tailoring these services to the user’s need is called adaptation or personalization. The in-formation which is used for making this adaptation or personalization is called the user model; alternatively, we could say that knowledge representation in the user model is used for adaptation. This information can be gathered by observing the user’s interaction with the system, by asking the user directly, and from the user’s profile. There are various adaptive hyper-media systems, such as content-based recommender systems, case-based recommender systems, personal-ized e-commerce applications, and adaptive news access systems. The systems which are used for edu-cational purposes are called Adaptive Educational

Hypermedia Systems. “User model” is an umbrella term under which the student model falls.

A good learning experience is one in which a learner can acquire new knowledge and skills, criti-cize any examined assumptions and beliefs, and en-gage in an energizing, collaborative search for knowledge and holistic personal development [91]. A User Model is: “a knowledge source, which contains explicit information on all aspects of the user that may be relevant to the dialog behaviors of the sys-tem” [92, p.6]. Personalization leads to more efficient learning [93–95]. Many researchers have discussed personalization in the systems described by them [96, 76, 97, 98]. Some researchers [99, 100] have de-scribed techniques of personalization. Access to Web-resources, especially learning objects, for use as supplementary material can be greatly facilitated by user models that describe an individual in terms of parameters such as name, age, grade of study, lan-guages known, level of academic achievement, and preferences for types of learning material. Web-based tutoring systems like ActiveMath [22] (for teaching mathematics), ELM-ART[101] (for providing adap-tive intelligent tutoring of material from an interac-tive online textbook), KBS Hyperbook [102] (an adaptive intelligent tutoring system for programming), Slide Tutor [103] (an intelligent tutoring system for visual classification problem solving), and SQL Tutor [36], are some examples of AEHSs.

As there was no formal way of representation for AEHSs in the beginning, [6] proposed some formal definitions: “An Adaptive Educational Hypermedia System (AEHS) is a quadruple (DOCS, UM, OBS, AC) with DOCS: Document Space: A finite set of first order logic (FOL) sentences with atoms for describ-ing documents (and knowledge topics), and predi-cates for defining relations between these documents. UM: User Model: A finite set of FOL sentences with atoms for describing individual users (user groups), and user characteristics, as well as predicates and rules for expressing whether a characteristic applies to a user. OBS: Observations: A finite set of FOL sentences with atoms for describing observations and predicates for relating users, documents/topics, and observations. AC: Adaptive Component: A finite set of FOL sentences with rules for describing adaptive functionality.” For example, NetCoach [104] pro-vides a framework for building adaptive hypermedia systems. It uses a knowledge base which consists of concepts, and the concepts are internal representa-tions of pages that can be presented to the learner. This is the basis for adaptive support in NetCoach. By using this, authors can create content-specific

relations between concepts and the knowledge base. The understanding for describing all the four parts has been developed from [6].

4.1. Document Space

The document space may include domain ontolo-gies (with domain concepts as atoms and connections between those concepts as predicates), a set of docu-ments and hierarchical connections between docu-ments, a set of tests to be offered, and relations be-tween test and document or test and concepts. For instance, in a general and very simple Adaptive Edu-cational Hypermedia System, the document space can be combination of a set of documents D = {D1, D2,……, Dk} and the relations among the documents are as: pre-requisite relation, contained_in or part_of relation. Pre-requisite relation means: it is a set of all documents Dj that are needed to be studied before another document Di and is writ-ten as: pre_requisite(Dj, Di) for some Di ≠ Dj

4.2. Observations

Observations contain information related to track-ing the user’s interaction with the system. For exam-ple, if there is a set of users say U = { U1, U2, …….., Uk} and an ob-servation of set of users who has read a particular document Dm is written as: observation(Dm, Ul, has_read) for for some Dm and Ul

4.3. User Models

User models draw conclusions about characteris-tics of users and what users have learned after ob-serving user interaction. An example of a simple user model is: a characteristic say ‘studied’ that is observed when user interact-ed with the system, and it can be written as: ∀Ui ∀Dl if [∃ Dt pre_requisite(Dt, Dl) ∧observation(Dt, Ui, has_read) ⇒ user_observation(Dl, Ui, Studied)

4.4. Adaptation Component

Adaptation component is a set of rules which are used at run time to provide feedback to students ac-cording to the information gathered from their user model. For the rule to state of the study of a document, an adaptive component used can be say suggested_for_reading and the traffic signal analogy for showing the adaptive feedback can be say red_flag and green_flag, can be written as: ∀Uk ∀Dl (∀Dmpre_requisite(Dl, Dm) ⇒ observation(Dm, Uk, Studied) ⇒ state_of_the_study (Dl, Uk, Suggested_for_reading)

Similar notations with more sets of relations, obser-vations, inference rules for user models, and adapta-tion components are used in NetCoach [104], ELM-ART [101], and KBS Hyperbook [102], which are described in detail in [6]. One of the major criticisms of [6]’s definition of AEHS was that it is difficult to re-use due to its usage of rules and logical framework at the higher level and its excessive domain depend-ence [105, 106].

5. Use of Semantic Web concepts in AEHSs

From the above review of four components in the definition of AEHS, it is clear that the key to adapta-tion is the knowledge model created or used at the back end, and the rules and reasoning used for ma-nipulating that knowledge model. An AEHS is an interconnection of two hyperspaces: knowledge space (i.e. concept-based hyperspace) and content-based hyperspace [107]. The document space can easily be replaced by ontology.

Ontology has its own knowledge representation, navigational capability, and the ability to be used for machine learning, and hence can be considered as part of concept-based hyperspace. Ontology is not only about the hierarchical connection of concepts: it also includes the OWL, description logic, and the inference rules. Hence the same knowledge can be used for the observations, user models, and adapta-tion component. The combined use of domain ontol-ogy, goal ontology, user model ontology and logical reasoning to create an adaptive web is argued for in [108]; the same can be applied in the case of web-based intelligent tutoring systems by using various types of ontologies. In Figure 7, an excerpt of appli-cation domain ontology for Java e-lecture is shown, which is taken from [108, p. 708]. Metadata stand-ards offer description of Learning Objects available on the Web, and ontology offers a way to describe domain knowledge and the learner. Due to RDF and the use of annotation properties, we can connect this ontological knowledge model to the educational con-tent on the Web, which is termed as a content-based hyperspace. For creating online courses and sequenc-ing the exercises in them, metadata of the course ma-terial, which is available in the form of LOs, can be used.

Fig. 7: An excerpt of application domain ontology for Java e-lecture [107, p. 708]

For distinguishing various kinds of information on a user/student (such as personal relations, security, preferences, performance, and portfolio information), a learner modeling standard is used, developed by IEEE, named Personal And Private Information (PAPI)26. A combination of RDF, RSFS and stand-ards like PAPI is used for applying personalization in web-based tutoring systems [109]. Ontological lan-guages use terminology from different namespaces; this gives scalability and interoperability to ontolo-gies. Many name spaces come from metadata stand-ards such as Dublin Core (dc) and various other standards such as RDF Schema (rdfs) (as explained in section 3.3). These can be used for resource annota-tions in the ontology, and thus can be used to link the online resource with its metadata file. An RDF in-stance of performance security record for the Ad-vanced Security Technologies for Networking II course, using PAPI standard, is shown in Figure 8, which is taken from [109].

Fig. 8: RDF instance of performance of security Record [106]

The use of RDF annotation can be applied for adap-tive course generation on the Web. The use of such

26 http://www.cen-ltso.net/Main.aspx?put=230

combinations of concepts in Semantic Web for creat-ing AEHS has been demonstrated by [110, 55]. The use of RDF and metadata in AEHS for sequencing and reuse of course material has been demonstrated and discussed in [97, 9]. The use of ontologies in AEHS has been discussed in [10, 96, 111]. TM4L [86] is a Topic Map-based course authoring adaptive hypermedia system in which concepts like XML, Ontology, RDF, and metadata are used. Topic Maps are also used as a tool for distributed learning envi-ronments based on the Web [112]. OntoAIMS [21] is a course authoring system in which the author can create course ontology, subject-domain ontology, and the educational metadata for the resources that will be attached as the course material; the same can be used for collaborative course authoring. OWL-OLM [20] is a framework for capturing the user’s conceptuali-zations based on dialogue generated by using ontolo-gy for implementation of effectual personalization in the Semantic Web. OWL-OLM is used by OntoAIMS for capturing the learner’s conceptualiza-tion by using the domain ontology. By OWL-OLM, the student’s interaction with the system (in the form of a graphical structure presented for each concept) is captured as a short-term knowledge model, and then the same is used for updating the long-term concep-tual model of the student.

Fig. 9: The LAOS model [94]

A framework named LAOS (Layered AHS Au-thoring-Model and Operators) (please see Figure 9), based on Semantic Web concepts, has been devel-oped by [106] for Adaptive Hypermedia Systems (AHSs), and is discussed in [65]. LAOS model con-tain five layers, namely: Domain Model (DM), Goal and Constraints Model (GM), User Model (UM), Adaptation Model (AM) and Presentation Model (PM). Domain model contains linked resources; with respect to educational system, it will have linked concepts and learning material. Goal model contains goal-related information like instructional strategies or pedagogic information about the resources. User model contains user-related information such as in-formation about student’s knowledge. Adaptation model contains the feedback specific to user/learner needs. Presentation model contains the display in-formation (such as color scheme) used for presenta-tion of the content. Presentation maps of all the five layers (i.e. domainMap, goalAndConstraintsMap, userMap, presentationMap, and adaptationStrategyMap) are defined formally [65]. MOT (My Online Teacher) [113] (discussed in sec-tion 7) is an adaptive educational hypermedia author-ing system based on LAOS framework. The key point in LAOS is that it is formulated with a view to separating the content from the authoring tool so that it will offer flexibility, expressivity, reusability, in-teroperability, non-redundancy, cooperation and standardization, which are overlapping goals of Se-mantic Web technology [106].

For already existing AEHS, semantic integration of such systems is proposed and discussed in [114]. For effective implementation of personalization in integrated systems one needs to translate the user models. Multiple translations can be avoided by using integration of user information collected on the fly, which is also called evidence [114]. Several problems are inherent in integration, such as naming conflicts, difference of graph structure, difference of scope, difference of granularity, and difference of focus. Use of ontology solves these problems. Use of single-ontology integration and central-ontology integration are the two approaches discussed by [114]. Single-ontology integration [114] means the use of common domain ontology as domain models for two systems so that their user models, which rely on the domain models, can be integrated easily. The integrations of Problets27 and QuizJET28

27

, and of SQL Tutor [36] and SQL Guide, are described in [114] to illustrate single-

http://www.problets.org 28 http://adapt2.sis.pitt.edu/quizjet/help.htm

ontology integration and central-ontology integration respectively. For these integrations two types of ar-chitectures are used: ADAPT2 (Advanced Distributed Architecture for Personalized Teaching & Training) [115], which is a framework for integrating adaptive and non-adaptive tools into distributed learning envi-ronments; and CUMULATE [116], which is a cen-tralized user modeling server. (More discussion of these two can be found in [115, 116, 117].) An OWL-based General User Model Ontology (GUMO) is also proposed and discussed by [118] for uniform interop-eration among distributed user models.

The next section discusses the application of Se-mantic Web concepts by the AI-ED research com-munity.

6. Use of Semantic Web concepts in AI-ED research

AIED research includes research in AI, Instruc-tional Design Science (IDS), and Learning Science (LS) [55]. Application of AI to Web-based ITS is very important because it offers many ways and tech-niques to implement intelligence via personalization, collaboration and adaptation in the ITS [56]. The issues found by the AI-ED research community were largely related to content. While analyzing the current state of the art of AI-ED research, [55] found several problems regarding authoring tools of intelligent tu-toring systems. Some of the issues are: the huge con-ceptual gap between authors and authoring systems; limited use due to lack of reusability and interopera-bility; lack of user-friendly interface for authoring ITS; lack of awareness of research in Information Science and learning theories by ITS authoring sys-tems. Considering the advantages of ontology, [55] proposed the use of ontological engineering for over-coming the problems stated above, and identified a need for extracting ontology from instructional theo-ries and instructional design models available in the literature. The notion of ontology engineering is de-scribed by [119] as: ontology learning or automat-ic/semi-automatic extraction of ontology from textu-al/non-textual sources by a software system. It uses NLP-based techniques, statistical techniques, and machine learning techniques for extracting concepts, attributes, instances, and axioms from textual and non-textual sources. Domain ontology engineering tools and techniques from the literature are surveyed in [119], which also identifies the need for reusable services for ontology learning, editing, merging and evaluation. One of the major limitations of techniques

of automatic extraction is that the entries produced can be inconsistent.

A blended learning approach, in which both hu-mans and computers are involved in the ontology engineering, is proposed in [55, 119]. As per the modified definition of ontology, it is a collaborative effort. Hence, for using ontology in ITS one needs to integrate collaborative authoring of ontology into the authoring of ITS [120]. The use of XML and RDF is also proposed for the effective use of ontology and connecting the ITS with the Web. Some researchers [59] have offered support framework based on ontol-ogy for authoring intelligent tutoring systems. The knowledge is divided into two parts: static and dy-namic. The static knowledge is about the curriculum instructions, and dynamic knowledge is generated from the adaptive component. The proposed frame-work includes the capacity to author a domain ontol-ogy, task ontology, and instructional design ontology. A description of the tasks as mentioned in the defini-tion of ATO (authoring task ontology) by [121] is shown in table 2.

Table 2

Definition and Description of Tasks in ATO

To integrate learning design standards and connect

those with theory, [58] described a formal process, which is closely parallel to the formal methods of creating ontology as described by [122-124]. The methodology is as follows: a) analysis of the domain; b) conceptualization, i.e. creating general classes and sub-classes and properties or relations among those classes; c) formalization; d) evaluation; and e) docu-mentation. This reflects the true value and advantage of the use of ontology, because both the processes were discovered by researchers in different fields and at different times. It also shows that the creation of ontology is a collaborative venture which should be agreed upon, and reveals that the evolved definition of ontology is a formal specification of shared con-ceptualization. Creating a domain model or expert module of any ITS is a time-consuming task. To save time for the creation of a domain model in a con-straint-based tutor, [14] used ontology to create a domain model automatically by extracting semantic and syntactic constraints from a domain ontology created by experts. Use of pedagogical agents in col-laboration with educational servers and ontologies is

one of the major contributions to AI-ED research achieved by the application of Semantic Web con-cepts to web-based ITS [56, 57]. The AI-ED research community is ready to accept the implementation of techniques of recommender systems as advancements in the ITS, which directs towards use of Semantic Web concepts and personalization. While discussing the development of an ontological web portal for ed-ucational ontologies, [125] presented a state of the art ontology of ontological technologies for education (as shown in Figure 10a & 10b), which gives us an overview of how to use ontology and the technolo-gies involved in developing educational applications. This state of the art ontology is divided into two parts: using ontologies and building ontologies. Us-ing ontologies (Figure 10a) include the application perspective of ontology and the technological per-spective, i.e. technologies in which ontologies are used. Ontologies are applied as various types of knowledge and as a cognitive tool. Ontologies are used in various technologies like Knowledge Repre-sentation, Information Retrieval and Semantic Web technology. Building ontologies (Figure 10b) include theoretical issues in ontology engineering and devel-opment of ontologies. These are generated manually and/or by machine using automatic and or semi-automatic methods.

Fig. 10a: State of the art Ontology of ontological technologies for education [125]

For automatic or semi-automatic generation of ontol-ogies, machine learning aspects, information retrieval aspects and some specific aspects of formal and semi-formal domains are used. For the development of ontologies there are various standards, languages and tools, which contain editors as well as ontology engi-neering environments. In the manual development of ontology, one faces subjective biases, pedagogical issues, and constraints; however, an ontology can be created collaboratively by considering concepts and

relationships among concepts, and writing the infer-ence rules to extract knowledge from ontologies.

Fig. 10b: State of the art Ontology of ontological technologies for education [125]

For creating this ontological web portal, the authors used Topic Map technology as it has a simple formal-ization of ontology that can be easily used for knowledge structuring and representation. A similar use of Topic Map technology as an educational re-source broker system for collaborative sharing of knowledge-centric content is discussed and demon-strated in [63].

The ITS research community has paid more atten-tion to student modeling than to modeling the tutor and tutoring strategies. The AI-ED research commu-nity considers this to be the major gap in ITS research. Work done in the field of expert systems has been seen as an effort to model the expert; maintaining a knowledge base for this purpose has been found to be a challenging task. The use of expert system technol-ogy for developing ITS is criticized by AI-ED re-searchers [126] for ignoring pedagogical theory or learning science theory and for not justifying the rules used for knowledge modeling. This was the major motivation behind the development of OMNI-BUS (Please see Figure 11), which is an ontology of learning theories and instructional theories that mod-els the instructional activities openly or in a hidden manner.

Learning theories are concerned with conditions under which learning happens and the way in which it happens, whereas instructional theories are con-cerned with ways to facilitate learning and maximize its outcome, that is, with best learning. The OMNI-BUS ontology is about target-world modeling, which contains an is-a hierarchy of not only the target con-cepts but also the states, actions, processes, events, and learning and instructional theories.

Fig. 11: Part of OMNIBUS Ontology

The change of state is defined as actions. Set of se-quential and hierarchical actions is defined as pro-cesses, and set of processes is defined as events. Edu-cational events are divided into two types: learning events and instructional events. In a learning event, the learner performs certain actions under certain conditions based on learning theories that have been designed to change her/his state of learning; whereas, in an instructional event, the instructor performs in-structional actions to facilitate learning actions that are hoped to eventually bring about a planned change in the learner’s learning state. In OMNIBUS, this kind of nested structure of learning and instructional events is defined as Instructional Learning Event (I_Levent). Further I_Levents are divided into three types of sub-events: simple event, reciprocal event (opposite of the simple event) and influential event. The influential event is further divided into an in-structional event, an effective learning event, and a prepared learning event. The instructional event in-fluences learning actions, thus causing an effective learning event (learning effect), which in turn pre-pares the base event (prepared learning event) for the

next learning action, so bringing about the required state of learning. From such a set of I_Levents an Instructional Learning scenario can be produced. By using OMNIBUS, an advanced authoring system named SMARTIES 29

In the next section the intelligent tutoring systems in which semantic web concepts are used are dis-cussed in brief.

is implemented (discussed in [126]).

7. Discussion of the ITS that Apply Semantic Web Concepts

Although many researchers have separately used Semantic Web concepts for educational applications, there exist very few Intelligent Tutoring Systems which have applied concepts of Semantic Web. ActiveMath [22] is a web-based intelligent tutoring system for teaching mathematics with a focus on high school and college level mathematics; it offers an interactive and exploratory learning environment to learners. ActiveMath30

SlideTutor [103] is an intelligent tutoring system for microscopic diagnosis of inflammatory diseases of the skin. It uses some architectural aspects of cog-nitive tutors, but also uses various types of ontologies such as domain ontology, domain task ontology, on-tology of abstract problem solving methods, and ped-agogical ontology. The abstract problem solving methods used in the expert model produce a dynamic solution graph with which the student interacts. Er-rors and hints are modeled on the instructional mod-ule (i.e. on pedagogical ontology) to provide person-alized training. SlideTutor

uses ontology for knowledge representation as well as for annotation of LOs, which contributes to re-usability and interoperability of the content. It is compliant with Dublin Core metadata standard, and uses standards like OpenMath, MathML (Markup language for mathematical ele-ments in the form of XML), and IEEELOM. It also provides various kinds of exercises and a user-friendly environment for practicing exercises. ActiveMath offers learners personalized feedback and the opportunity to create their own personalized hyper-book for learning various concepts. It is one of the intelligent tutoring systems that have adaptive hypermedia.

31

29

is a system which is used

http://www.ei.sanken.osakau.ac.jp/pub/miz/AIED07WS_Invited.pps

30 http://www.activemath.org 31 http://slidetutor.upmc.edu

for visual classification problem solving, and is based on the information processing theory of cognition. Because of the use of ontologies, SlideTutor offers flexibility in the instructional layer, scalability, reus-ability, and easy maintenance. The dynamic solution graph offers backward and forward reasoning. For creating ontologies in SlideTutor, Protégé is used, and production rules that are used for manipulating the ontologies are written in Jess32

MOT (My Online Tutor) [113] is a web-based au-thoring AHS environment based on LAOS, a five-layer framework discussed in section 5. It uses con-cept map-based domain knowledge at the front end and XML-based format at the back end, which offers it flexibility and reusability in the Web-based envi-ronment. Using MOT, users can create their own tu-toring courses or modules. The use of LAG, the adap-tation language, offers the greater level of semantics that is discussed in [127]. For converting MOT into an AEHS, common adaptation format (CAF) is used, whose schematic representation is similar to the LAOS XML schema. In MOT, Semantic Web no-tions such as XML, RDF, RDF Schema, LOs and metadata are used. The tutoring courses or modules authored by using MOT can offer intelligent tutoring because of personalization.

, a Java production rule system. SlideTutor is an intelligent tutoring sys-tem based on the expert system model.

SMARTIES is an innovative intelligent tutor au-thoring system that is aware of learning and instruc-tional theories based on OMNIBUS. Basically OM-NIBUS is about “what to learn?” and “how to learn?” (as described at the end of section 6). OMNIBUS provides basic building blocks to SMARTIES for creating instructional learning scenarios in collabora-tion with the author, i.e. the instructional designer (ID). Decomposition of I_Levent is central to SMARTIES. The I_Levent can be considered an in-structional strategy used by the author (ID). This de-composition can be done in two ways. The first way is to explain what to learn and how to learn. The se-cond way is just to display examples or content and let the learner to explore learning. The decomposition can be proposed to the author (ID) by the system based on OMNIBUS ontology, and then the author can select I_Levent components from the information available. For example, if a learner needs help, then way1 tells him/her what to learn by suggesting a hy-pothetical experiment; then, following the construc-tivist theory of learning, the system suggests an actu-al interactive experiment. Way2 would suggest

32 http://herzberg.ca.sandia.gov/jess/

providing related content and examples to display. The author (ID) could use his/her own instructional strategies to create an I_Levent and to produce a learning scenario. The instructional learning scenari-os produced by SMARTIES are IMS-LD33-compliant and can be connected to Learning Objects, such as virtual lab instruments available through GLOBE 34

Table 3 shows a classification of the four systems as per the usage of Semantic Web concepts.

(Global Learning Objects Brokered Exchange).

Table 3

Classification of the systems based on usage of Semantic Web concepts

ITS/Authoring Systems

Semantic Web Concepts RDF/XML Ontology LOs/Metadata

ActiveMath Yes Yes Yes SlideTutor Yes MOT Yes Yes SMARTIES Yes Yes Yes

It is clear from the above table that maximum appli-cation of Semantic Web concepts has been made by ActiveMath. The base of ontology editor Protégé is in expert systems. SlideTutor is a kind of expert system which provides personalized feedback. The system is very domain-specific, and confined to a small part of the domain, i.e. tutoring of medical diagnosis by us-ing visual slides. Because of this limitation dictated by the needs of the tutoring domain, other Semantic Web concepts such as RDF/XML and LOs/Metadata are not used in that system. As per the intelligent tu-tor authoring tools, MOT comes from AEHS back-ground and does not show any evidence of the use of ontology; however, there is a potential to use the on-tology in MOT. SMARTIES come from AI-ED background, and apply all the concepts of Semantic Web. It is mentioned clearly in [126] that although SMARTIES looks like an expert system, it actually is not: it shows personalization in instructional strate-gies, but does not explore personalization of the learner by using information about the learner.

8. Conclusion

Reviewing the literature from ITS, AEHS and the AI-ED community, it becomes clear that all the three fields are moving in parallel, but have little intercon-nection and collaboration. Notions of Semantic Web like RDF/XML, LOs, and Metadata are building

33 http://www.imsglobal.org/learningdesign/ldv1p0.

html 34 http://www.globe-info.org/

blocks that help to use ontology effectively. Student modeling techniques offered by ITS, personalization offered by AEHS, and the modeling of instructional strategies offered by the AI-ED community would add greater intelligence if used in combination with Semantic Web concepts. To combine the work in these three fields and enhance the intelligence of the tutoring systems, effective use of Semantic Web con-cepts in combination with some AI techniques such as ontology engineering would help. Also, for exist-ing stand-alone AEHS, a way to increase scalability and re-usability would be to use a central ontological server in combination with annotation of LOs and metadata.

Acknowledgement

The research project for which this literature sur-vey work has been done is funded by HP Labs, India and International Institute of Information Technology, Bangalore. We thank both the organizations for ex-tending all their infrastructural support, including library resources.

References

[1] P. A. Cohen, J. A. Kulik & C. C. Kulik, Educational Out-comes of Tutoring: A meta-analysis of findings, American Educational Research Journal, Vol. 19,pp. 237-248.

[2] A. T. Corbett, K. R. Koedinger & J. R. Anderson, Intelligent Tutoring Systems, In M. Helander, T. K. Landauer, P. Prabhu (Eds.): Handbook of Human-Computer Interaction, Second, Completely Revised Edition Elsevier Science B. V., pp. 849-874, 1997.

[3] P. Brusilovsky, The Construction and Application of Student Models in Intelligent Tutoring Systems, Journal of Computer and Systems Sciences International, Vol. 32 (1), pp. 70-89, 1994.

[4] T. Murray, Authoring Intelligent Tutoring Systems: An Anal-ysis of the State of the Art, International Journal of Artificial Intelligence in Education, Vol. 10, pp. 98-129, 1999.

[5] P. Brusilovsky, Adaptive Hypermedia, User Modeling and User-Adapted Interaction, Vol.11, Kluwer Academic Publish-er, pp. 87-110, 2001.

[6] N. Henze & W. Nejdl, Logically Characterizing Adaptive Educational Hypermedia Systems, In International Workshop on Adaptive Hypermedia and Adaptive Web-based Systems (AH2003), Budapest, Hungary, 2003.

[7] W. Nejdil & M. Wolpers, KBS Hyperbook – A Data-Driven Information System on the Web, In Eighth International World Wide Web Conference(WWW1999), Toronto, Canada, 1999.

[8] P. Karampiperis & D. Sampson, Adaptive Learning Resources Sequencing in Educational Hypermedia Systems, Educational Technology & Society, Vol. 8 (4), pp. 128-147, 2005.

[9] A. Papasalouros; S. Retails & N. Papaspyrou, Semantic De-scription of Educational Adaptive Hypermedia based on a Conceptual Model, Educational Technology & Society, Vol. 7 (4), pp. 129-142, 2004.

[10] P. Brusilovsky & C. Peylo, Adaptive and Intelligent Web-based Educational Systems, International Journal of Artificial Intelligence in Education, Vol. 13, pp. 156-169, 2003.

[11] S. A. Mcllraith, T. C. Son & H. Zeng, Semantic Web Services, IEEE Intelligent Systems, Vol.16(2), pp. 46-53, 2001.

[12] M. Tallis, N. Goldman & R. Balzer, The Briefing Associate: A Role for COTS Applications in the Semantic Web, SWWA’01: The First Semantic Web Working Symposium, 2001.

[13] P. Dolog, R. Gavriloaie, W. Nejdl & J. Brase, Integrating Adaptive Hypermedia Techniques and Open RDF-based Envi-ronments, WWW2003, May 20-24, Budapest, Hungary, 2003.

[14] P. Suraweera, A. Mitrovic & B. Martin, The use of ontologies in ITS domain knowledge authoring, In: Mostow,J. and Tedesco, P. (eds.) Proceedings of 2nd International Workshop on Applications of Semantic Web for E-Learning SWEL’04, ITS2004, Maceio, Brazil, pp. 41-49, 2004.

[15] CH. Lu, SH. Wu, LY. Tu & WL. Hsu, The Design of an Intel-ligent Tutoring System Based on the Ontology of Procedural Knowledge, In Proceedings of the IEEE International Confer-ence on Advanced Learning Technologies (ICALT’04), Joensuu, Finland, pp. 525-529, 2004.

[16] A. Mitrovic & V. Devedzic, A model of multitutor ontology-based learning environments, International Journal of Contin-uing Engineering Education and Lifelong Learning, Vol. 14 (3), pp. 229- 245, 2004.

[17] W. Huang, D. Webster, D. Wood & T. Ishaya, An intelligent semantic e-learning framework using context-aware semantic web technologies, British Journal of Educational Technology, Vol. 37 (3), pp. 351–373, 2006.

[18] A. Canales, A. Pena, R. Peredo, H. Sossa & A. Gutierrez, Adaptive and intelligent web based education system: To-wards an integral architecture and framework, Expert Systems with Applications, Vol. 33, pp. 1076-1089, 2007.

[19] J. Clemente, J. Ramirez & A. Antonio, A Proposal for student modeling based on ontologies and diagnosis rules, Expert Sys-tems with Applications, Vol. 38, pp. 8066-8078, 2011.

[20] R. Denaux, L. Aroyo & V. Dimitrova, OWL-OLM: Interactive Ontology-based Elicitation of User Models, In: Proceedings of Workshop on Personalisation for the Semantic Web (PerSWeb 2005) at UM 2005, Edinburgh, UK, July 23-29, 2005.

http://www.win.tue.nl/persweb/full-proceedings.pdf [21] L. Aroyo & D. Dicheva, AIMS: Learning and Teaching Sup-

port for WWW-based Education, International Journal for Continuing Engineering Education and Life-long Learning (IJCEELL), 11(1/2), 152–164, 2001.

[22] E. Melis, E. Andres, J. Budenbender, A. Frischauf, G. Goguadze, P. Libbrecht, M. Pollet & C. Ullrich, ActiveMath: A Generic and Adaptive Web-based Learning Environment, International Journal of Artificial Intelligence in Education, Vol. 12, pp. 385-407, 2001.

[23] E. Wenger, Artificial Intelligence and Tutoring Systems, Los Altos, CA: Morgan Kaufmann, 1987.

[24] S. Ohlsson, Some Principles of Intelligent Tutoring, In Lawler & Yazdani (Eds.): Artificial Intelligence and Education, Vol. 1, Ablex: Norwood, NJ, pp. 203-238, 1987.

[25] V. Shute & J. Psotka, Intelligent Tutoring Systems: Past, pre-sent and future. In D. Jonassen (Eds.): Handbook of Research for Educational Communications and Technology, New York: Macmillan, pp. 570-600, 1996.

[26] H. L. Burns & C. G. Capps, Foundations of Intelligent Tutor-ing Systems: An Introduction, In J. J. Richardson & M. C. Polson (Eds.): Proceedings of the Air Force Forum for Intelli-

gent Tutoring Systems, Air Force System Command, Brooks Air Force Base, Texas, pp. 1-14, 1989.

[27] J. R. Anderson; A. T. Corbett; K. R. Koedinger & R. Pelletier Cognitive Tutors: Lessons Learned, The Journal of the Learn-ing Sciences, Vol. 4 (2), pp. 167-207, 1995.

[28] J. R. Anderson, Rules of the Mind. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc., 1993.

[29] J. G. Greeno & H. A. Simon, Problem solving and reasoning, In R. C. Atkinson, R. Hernstein, G. Lindzey and R. D. Luce (Eds.): Steven’s handbook of experimental psychology, New York: John Wiley and Sons, 1988.

[30] B. P. Woolf, Student Modeling, In R. Nkambou, J. Bourdeau & R. Mizoguchi (Eds.): Advances in Intelligent Tutoring Sys-tems, Springer – Verlag Berlin Haidelberg, pp. 267-279, 2010.

[31] K. VamLehn, Student Modeling, In J. J. Richardson & M. C. Polson (Eds.): Proceedings of the Air Force Forum for Intelli-gent Tutoring Systems, Air Force System Command, Brooks Air Force Base, Texas, pp. 49-71, 1989.

[32] B. Woolf, Building Intelligent Interactive Tutors: Student-Centered Strategies for Revolutionizing E-learning, San Fran-cisco, CA: Elsevier Inc., Morgan Kauffman, 2009.

[33] K. VanLehn, C. Lynch, K. Schulze, J. A. Shapiro, R. Shelby, L. Taylor, D. Treacy, A. Weinstein & M. Wintersgill, The An-des physics tutoring system: Lessons learned, International Journal of Artificial Intelligence in Education, Vol. 15(3), pp.147–204, 2005.

[34] S. B. Blessings, A Programming by Demonstration Authoring Tool for model-tracing tutors, International Journal of Artifi-cial Intelligence in Education, Vol. 8, pp. 233-261, 1997.

[35] S. Ohlsson, Learning from Performance Errors, Psychological Review, Vol. 102, pp. 241-262, 1996.

[36] A., Mitrovic & K. Hausler, Porting SQL-Tutor to the Web, Proc. ITS’2000 workshop on Adaptive and Intelligent Web-based Education Systems, pp. 37-44, 2000.

[37] B. Martin & A. Mitrovic, WETAS: a Web-Based Authoring System for Constraint-Based ITS, Proc. 2nd Int. Conf on Adaptive Hypermedia and Adaptive Web-based Systems AH 2002, Malaga, Spain, LCNS, pp. 543-546, 2002.

[38] P. Suraweera & A. Mitrovic, KERMIT: A Constraint-based Tutor for Database Modeling, In S.A. Cerri, G. Gouardères, and F. Paraguaçu (Eds.): ITS 2002, LNCS 2363, Springer-Verlag Berlin Heidelberg, pp. 377–387, 2002.

[39] A. Mitrovic, K. R. Koedinger & B. Martin, A Comparative Analysis of Cognitive Tutoring and Constraint-Based Model-ling, In P. Brusilovsky, A. Corbett & F. D. Rosis (Eds.): Pro-ceedings of the 9th International Conference on User Modeling UM’2003, Vol. LNAI 2702, Berlin: Springer-Verlag, pp. 313-322, 2003.

[40] M. Mayo & A. Matrovic, Optimizing ITS Behavior with Bayesian Networks and Decision Theory, International Jour-nal of Artificial Intelligence in Education, Vol. 12, pp. 124-153, 2001.

[41] P. Jackson, Introduction to Expert Systems (3rd ed), Addison Wesley, pp.2, 1998.

[42] T. A. Byrd & R. D. Hauser, Expert systems in production and operations management: research directions in assessing over-all impact, International Journal of Production Research, Vol. 29 (12), PP. 2471-2482, 1991.

[43] S. Andreassen, F. V. Jensen & K. G. Olsen, Medical expert systems based on causal probabilistic networks, International Journal of Biomedical Computing, Vol. 28, pp. 1-30, 1991.

[44] F. A. Mora, G. Passariello, G. Carrault, & J. L. Pichon, Intelli-gent patient monitoring and management systems: A review, IEEE Engineering in Medicine and Biology, pp. 23-33, 1993.

[45] B. Chandrasekaran & S. Mittal, Deep versus compiled knowledge approaches to diagnostic problem solving, Interna-

tional Journal of Man-Machine Studies, Vol. 19, pp. 425-436, 1983.

[46] B. Chandrasekaran, Towards a taxonomy of problem solving types, AI Magazine, Vol. 4 (1), pp. 9-17, 1983.

[47] B. Chandrasekaran, Generic tasks in knowledge-based reason-ing: High-level building blocks for expert system design, IEEE Expert, Vol. 1(3), pp. 23-30, 1986.

[48] B. Chandrasekaran, T. R. Johnson & J. W. Smith, Task-structure analysis for knowledge modeling, ACM 35, pp. 124-136, 1992.

[49] B. Chandrasekaran & J. R. Josephson, Ontologies: What are they? why do we need them?. IEEE Intelligent Systems and their Applications, Vol. 14(1), pp. 20-26, 1999.

[50] M. A. Musen, Use of a domain model to derive an interactive knowledge-editing tool. International Journal of Man-Machine Studies, Vol. 26, pp. 105-121, 1987.

[51] E. H. Shortliffe, D. P. Scott, M. B. Bischoff, A. B. Campbell, W. van Melle & C. D. Jecobs, ONCOCIN: An expert system for oncology protocol management, International Joint Con-ference on Artificial Intelligence (IJCAI’81), Vancouver, CA, pp. 876-881, 1981.

[52] J. H. Gennari, M. A. Musen, R. W. Fergerson, W. E. Grosso, M. Crubezy, H. Erikson, N. F. Noy & S. W. Tu, The evolution of protege: An environment for knowledge-based systems de-velopment, International Journal of Human-Computer Studies, Vol. 58 (1), pp. 89-123, 2003.

[53] J. Bourdeau & M. Grandbastein, Modeling Tutoring Knowledge, In R. Nkambou, J. Bourdeau & R. Mizoguchi (Eds.): Advances in Intelligent Tutoring Systems, Springer-Verlag Berlin Heidelberg, pp. 123-143, 2010.

[54] R. R. Burton, The Environment Module of Intelligent Tutoring Systems, In J. J. Richardson & M. C. Polson (Eds.): Proceed-ings of the Air Force Forum for Intelligent Tutoring Systems, Human Resource Laboratory, Air Force Systems Division, Texas, pp. 99-123, 1989.

[55] R. Mizoguchi & J. Bourdeau, Using Ontological Engineering to Overcome Common AI-ED Problems, International Jour-nal of Artificial Intelligence in Education, Vol. 11, pp. 107-121, 2000.

[56] V. Devedzic, Web Intelligence and Artificial Intelligence in Education, Educational Technology & Society, Vol. 7 (4), pp. 29-39, 2004.

[57] V. Devedzic, Education and the Semantic Web, International Journal of Artificial Intelligence in Education, Vol. 14, pp. 39-65, 2004.

[58] V. Psyche, J. Bourdeau, R. Nkambou & R. Mizoguchi, Mak-ing Learning Design Standards Work with an Ontology of Ed-ucational Theories, Proceedings of AIED2005, IOS Press, pp. 539-546, 2005.

[59] L. Aroyo & R. Mizoguchi, Authoring Support Framework for Intelligent Educational Systems, Proceedings of AIED2003, IOS Press, pp. 362-365.

[60] J. R. Miller, The Role of Human Computer Interaction in Intelligent Tutoring Systems, In In J. J. Richardson & M. C. Polson (Eds.): Proceedings of the Air Force Forum for Intelli-gent Tutoring Systems, Human Resource Laboratory, Air Force Systems Division, Texas, pp. 137-170, 1989.

[61] B. P. Woolf, I. Arroyo, D. Cooper, W. Burleson & K. Muldner, Affective Tutors: Automatic Detection of and Response to Student Emotion, In R. Nkambou, J. Bourdeau & R. Mizoguchi (Eds.): Advances in Intelligent Tutoring Systems, Springer-Verlag Berlin Heidelberg, pp. 207-227, 2010.

[62] C. Bizer, T. Heath & T. Berners-Lee, Linked Data – The Story So Far, In T. Heath, M. Hepp & C. Bizer (Eds.): Special Issue on Linked Data, International Journal on Semantic Web & In-formation Systems (IJSWIS), 2009.

[63] A. Lalingkar, C. Ramanathan & S. Ramani, An Educational Resources Broker System for Collaborative Sharing of Knowledge-Centric Content, In: Proceedings of IEEE Interna-tional Conference on Technology for Education T4E’11, Chennai, July 2011, pp.61-68, 2011.

[64] T. R. Gruber, A translation approach to portable ontologies, Known Acquisition, Vol. 5, No. 2, pp. 199-220, 1993.

[65] A. I. Cristea, What can the Semantic Web do for Adaptive Educational Hyeprmedia?, Educational Technology & Society, Vol. 7 (4), pp. 40-58, 2004.

[66] M. C. Daconta, L. J. Obrst & K. T. Smith, The Semantic Web: A Guide to the Future of XML, Web Services, and Knowledge Management, New York: Wiley, 2003.

[67] J. Hendler, Agents and the Semantic Web, IEEE Intelligent Systems, 16 (2), pp. 30-37, 2003.

[68] C. Brewster & K. O’Hara, Knowledge Representation with Ontologies: The Present and Future, IEEE Intelligent Systems, Vol. 19 (1), pp. 72-81, 2004.

[69] W. N. Borst, Construction of Engineering Ontologies, Un-published PhD thesis, University of Twente, Enschede, 1997.

[70] T. Passin, Explorer’s Guide to the Semantic Web, Manning Publications Co., 2004.

[71] L. M. Garshol, Metadata? Thesauri? Taxonomies? Topic Maps! Making sense of it all, Ontopia, 2004. http://www.ontopia.net/topicmaps/materials/tm-vs-thesauri.html

[72] V. Devedzic, Introduction to Web-based Education, in Seman-tic Web and Education, Springer-Science business Media, LLC, 2006.

[73] IEEE LTSC: IEEE Learning Technology Standards Commit-tee, Draft Standard for Learning Object Metadata Version 6.1, 2001. http://ltsc.ieee.org/doc/

[74] R. McGreal, Learning Objects: A Practical Definition, In D. Perrin, S. Downes, B. Muirhead & E. Perrin (Eds.) Proceed-ings of International Journal of Instructional Technology and Distance Learning, ISSN 1550-6908, 1 (9), pp. 21-32, 2004.

[75] P. R. Polsani, Use and Abuse of Reusable Learning Objects, Journal of Digital Information, 3 (4), 2003.

[76] P. Brusilovsky & H. Nijhawan, A Framework for Adaptive E-Learning Based on Distributed Re-usable Learning Activities, In M. Driscoll and T.C. Reeves (Eds.): Proceddings of World Conference on E-Learning, E-Learn 2002 Montreal, Canada, 2002.

[77] F. Neven & E. Duval, Reusable Learning Objects: A Survey of LOM-Based Repositories, Multimedia’02, ACM, December 1-6, pp. 291-294, 2002.

[78] P. Mohan & C. Brooks, Learning Objects on the Semantic Web, In Proceedings of CALT 2003, 2003.

[79] A. Lalingkar & S. Ramani, A Topic Map-based System for Identifying Relevant Learning Objects, In: Proceedings of World Conference on Educational Multimedia, Hypermedia and Telecommunications 2010, ED-MEDIA 2010, Chesapeake, VA: AACE, pp. 1044-1053, 2010.

[80] S. Pepper, The TAO of Topic Maps, In: Proceedings of XML Europe, 2000.

[81] ISO 13250, Guidelines for the establishment and development of monolingual thesauri (ISO, Geneva 1986), International Organization for Standardization, ISO 2788:1986.

[82] Ontopoly: The Topic Map Editor User’s Guide, Ontopia, 2009. http://omnigator.topicmapslab.de/ontopoly/doc/user-guide.html

[83] L. M. Garshol, Metadata? Thesauri? Taxonomies? Topic Maps! Making sense of it all, Ontopia, 2004. http://www.ontopia.net/topicmaps/materials/tm-vs-thesauri.html

[84] V. Devedzic, Introduction to Web-based Education, in Seman-tic Web and Education, Springer-Science business Media, LLC, 2006.

[85] T. Passin, Explorer’s Guide to the Semantic Web, Manning Publications Co., 2004.

[86] D. Dicheva & C. Dichev, TM4L: Creating and Browsing Educational Topic Maps, British Journal of Educational Technology - BJET, 37 (3), pp. 391-404, 2006.

[87] S. Lavik & T. W. Nordeng, BrainBank Learning –Building Topic Maps based e-portfolios, In A.J. Canas; J. D. Novak and F. M. Gonzalez (Eds): Proceedings of the first International Conference on Concept Mapping, Pamplona, Spain, 2004.

[88] G. Mihai, L. Stanescu, D. Burdescu & M. Brezovan, A Topic-Map for Subject-Centric Learning, In G. A. Papadopoulos & C. Badica (Eds.): Intelligent Distributed Computing III, SCI 237, Springer-Verlag Berlin Heidelberg, pp. 141-150, 2009.

[89] P. Brusilovsky, Methods and Techniques of Adaptive Hyper-media, User Modeling and User-Adapted Interaction, Vol. 6, pp. 87-129, 1996.

[90] D. Sampson, C. Karagiannidis & Kinshuk, Personalized Learning: Educational, Technological and Standardisation Perspective, Interactive Educational Multimedia, Vol. 4, ISSN 1576-49990, pp. 24-39, 2002.

[91] D. Eastmond & L. Ziegahn, Instructional design for the online classroom, In Z. L. Berge & M. P. Collins (Eds): Computer-mediated Communication and the Online Classroom, Vol. 3: Distance Education, Cresskill, NJ: Hampton Press, pp. 59-80, 1995.

[92] W. Wahlster & A. Kobsa, User Models in Dialog Systems, In A. Kobsa & W. Wahlster (Eds): User Models in Dialog Systems, Berlin: Springer, pp. 4-34, 1989.

[93] C. M. Chen, H. M. Lee & Y. H. Chen, Personalized e-Learning System using Item Response Theory, Computers & Education, Vol. 44, N. 3, pp. 237-255, 2005.

[94] C. M. Chen, Intelligent Web-based Learning System with Personalized Learning path Guidance, Computers & Educa-tion, Vol. 51, pp. 787-814, 2008.

[95] L. Aroyo & D. Dicheva, The New Challenges for e-Learning: The Educational Semantic Web, Educational Technology & Society, Vol. 7, No. 4, pp. 59-69, 2004.

[96] P. Brusilovsky & N. Henze, Open Corpus Adaptive Educa-tional Hypermedia, In P. Brusilovsky, A. Kobsa, and W. Nejdl (Eds.): The Adaptive Web, LNCS 4321, Springer-Verlag Ber-lin Heidelberg, pp. 671-696, 2007.

[97] P. Dolog & M. Sintek, Personalization in Distributed e-Learning Environments, WWW2004, May 17-22, New York, USA, pp.170-179, 2004.

[98] H. C. Muntean & G. M. Muntean, Open corpus architecture for personalized ubiquitous e-Learning, Personal and Ubiqui-tous Computing, Vol.13, Springer-Verlag London, pp. 197-205, 2009.

[99] N. Henze & W. Nejdl, Adaptation in Open Corpus Hyperme-dia, International Journal of Artificial Intelligence in Educa-tion, 12, pp. 325-350, 2001.

[100] D. Sampson, C. Karagiannidis & F. Cardinali, An Ar-chitecture for Web-based e-Learning Promoting Re-usable Adaptive Educational e-Content, Educational Technology & Society, Vol. 5, No. 4, pp. 27-37, 2002.

[101] G. Weber & P. Brusilovsky, ELM-ART: An adaptive versatile system for web-based instruction, International Journal of Artificial Intelligence in Education, Vol. 12, pp. 351-384, 2001.

[102] W. Nejdil & M. Wolpers, KBS Hyperbook – A Data-Driven Information System on the Web, In proceedings of 8th International Workshop on World Wide Web, Toronto, Canada, 1999.

[103] R. Crowley, O. Medvedeva & D. Jukic, SlideTutor: A model-tracing Intelligent Tutoring System for teaching micro-scopic diagnosis. In U. Hoppe, F. Verdejo, & J. Kay (Eds.): Artificial Intelligence in Education, Amsterdam: IOS Press/Tokyo: OHM Ohmsha, pp. 157-164, 2003.

[104] G. Weber, H.C. Kuhl & S. Weibelzahl, Developing adaptive internet based courses with the authoring system NetCoach, In Proceedings of the 3rd Workshop on Adaptive Hypermedia AH2001, Sonthofen, Germany, July 14, Technical University Eindhoven, pp. 35-48, 2001.

[105] A.I. Cristea, D. Floes, N. Stash & P. De Bra, MOT meets AHA!. In Proceedings of PEG’03 Conference (St. Pe-tersburg, Russia, 2003.

[106] A. Cristea & A. De Mooij, LAOS: Layered WWW AHS Authoring Model and its corresponding Algebraic Oper-ators. In Proceedings of World Wide Web’03, alternate educa-tion track, 20-24 May, Budapest, Hungary, 2003.

[107] P. Brusilovsky, Developing Adaptive Educational Hy-permedia Systems: from Design to Authoring Tools, In T. Murry, S. Blessings & S. Ainsworth (eds.) Authoring Tools for Advanced Technology Learning Environments: Toward Cost-Effective, Adaptive, Interactive and Intelligent Educational Software, Dordrecht, Kulwer, pp. 377-409, 2003.

[108] P. Dolog & W. Nejdl, Semantic Web Technologies for the Adaptive Web, In P. Brusilovskly, A. Kobsa & W. Nejdl (Eds.): The Adaptive Web, LNCS4321, pp. 697-719, 2007.

[109] P. Dolog & W. Nejdl, Challenges and benefits of the semantic Web for user modeling. In: De Bra, P., Davis, H., Kay, J. and schraefel, m. (Eds.): Proc. of AH2003: Workshop on adaptive hypermedia and adaptive Web-based systems, (Budapest, Hungary, Eindhoven University of Technology, pp. 99-111, 2003.

[110] N. Henze, P. Dolog & W. Nejdl, Reasoning and Ontol-ogies for Personalized e-Learning in the Semantic Web, Edu-cational Technology & Society, Vol. 7 (4), pp. 82-97, 2004.

[111] P. Dolog, N. Henze & W. Nejdl, Logic-based Open Hypermedia for the Semantic Web, In Proc. of International Workshop on Hypermedia and the Semantic Web, Hypertext 2003 Conference, Nottinngham, UK, 2003.

[112] L. Aroyo & D. Dicheva, The New Challenges for E-learning: The Educational Semantic Web. Educational Tech-nology & Society, Vol. 7 (4), pp. 59-69, 2004.

[113] A. Cristea & A. De Mooij, Adaptive Course Authoring: MOT, My Online Teacher. In the Proceedings of International Conference on Telecommunications (ICT2003), February 23 - March 1, 2003, Tahiti, Papeete - French Polynesia, 2003.

[114] S. Sosnovsky, P. Brusilovsky, M. Yudelson, A. Mitrovic, M. Mathews & A. Kumar, Semantic Integration of Adaptive Educational Systems, In T. Kuflik, S. Berkovsky, F. Carmagnola, D. Heckmann and A. Krüger (Eds.): Advances in Ubiquitous User Modelling, Springer, LNCS Vol. 5830, pp. 134-158, 2009.

[115] P. Brusilovsky, S. Sosnovsky & M. Yudelson, Ontolo-gy-based framework for user model interoperability in distrib-uted learning environments. In: Richards, G. (Ed.) Proceed-ings of World Conference on E-Learning, E-Learn 2005, Van-couver, Canada, October 24-28, pp. 2851–2855, 2005.

[116] M. Yudelson, P. Brusilovsky & V. Zadorozhny, A User Modeling Server for Contemporary Adaptive Hypermedia: An Evaluation of Push Approach to Evidence Propagation. In: Conati, C., McCoy, K., Paliouras, G. (Eds.): UM 2007. LNCS (LNAI), Springer, Heidelberg, Vol. 4511, pp. 27–36, 2007.

[117] P. Brusilovsky, Knowledge Tree: A Distributed Archi-tecture for Adaptive e-Learning, WWW 2004, May 17-22, New York, USA, ACM, pp. 104-113, 2004.

[118] D. Heckmann, T. Schwartz, B. Brandherm, M. Schmitz & M. Wilamowitz-Modellendorff, GUMO – The General User Model Ontology, In L. Ardissono, P. Brna & A. Mitrovic (Eds.): UM 2005, LNAI 3538, Springer–Verlag Berlin Heidel-berg, pp. 428-432, 2005.

[119] A. Zouaq & R. Nkambou, A Survey of Domain Ontol-ogy Engineering Methods and Tools, In In R. Nikambou, J. Bourdeau & R. Mizoguchi (Eds.): Advances in Intelligent Tu-toring Systems, SCI 308, Springer-Verlag Berlin Heidelberg, pp. 103-109, 2010.

[120] J. Bourdeau & R. Mizoguchi, Collaborative Ontological Engineering of Instructional Design Knowledge for an ITS Authoring Environment, In S. Cerri, G. Gouarderes, & F. Paraguacu (Eds.): Proceedings of The 6th International Con-ference on Intelligent Tutoring Systems ITS’2002, Biarritz, France and San Sebastian, Spain, pp. 399-409, 2002.

[121] R. Mizoguchi, K. Sinitsa & M. Ikeda, Task ontology design for intelligent educational/training systems, Workshop on Architectures and Methods for Designing Cost-Effective and Reusable ITSs. In Proc. of ITS’96 Conference, Montreal, Canada, 1996.

[122] M. Fernandez, A. Gomez-Perez, J. Pazos & A. Pazos, Ontology of tasks and methods, IEEE Intelligent Systems and their Applications, Vol. 14 (1), pp. 37-46, 1999.

[123] A. Gomez-Perez, Knowledge Sharing and Reuse. In J. Liebowitz (Ed): The Handbook of Applied Expert Systems, CRC, 1998.

[124] M. Uschold & M. Gruninger, Ontologies: principles, methods, and applications, Knowledge Engineering Review, Vol. 11 (2), pp. 93-155, 1996.

[125] D. Dicheva, S. Sonovsky, T. Gavrilova & P. Brusilovsky, Ontological Web Portal for Educational Ontolo-gies, In Proceedings of International Workshop on Applica-tions of Semantic Web Technologies for E-Learning (SW-EL 05), July 2005

[126] R. Mizoguchi, Y. Hayashi & J. Bourdeau, Ontology-based Formal modeling of the Pedagogical World: Tutor Modeling, In R. Nikambou, J. Bourdeau & R. Mizoguchi (Eds.): Advances in Intelligent Tutoring Systems, SCI 308, Springer-Verlag Berlin Heidelberg, pp. 229-247, 2010.

[127] A. Cristea, D. Smits & P. De Bra, Writing MOT, Read-ing AHA! – Converting between an authoring and a delivery system for adaptive educational hypermedia, In Proceedings of A3EH Workshop at the AIED’2005, Amsterdam, The Nether-lands, pp. 107 -114, 2005.