International Journal of Advances in Science and Technology, Vol. 3, No.

3, 2011

Ontology Mapping To Support E-Learning Resource Reuse

Tatyana Ivanova1
1

Technical University of Sofia, College of Energy and Electronics, Botevgrad, Bulgaria [email protected]

Abstract
Development of learning resources is expensive and time-consuming tack (especially for adaptive resources). That is why many efforts have been done to support Learning Objects (LO) reuse. To facilitate reusing, automatic comparison of Los on semantic level in one system and learning goals in other system (needed for resources) should be performed. This is possible only if learning management systems use semantically represented metadata and knowledge (as ontologies) and can make semantic mappings between these recourse and learning goal representations. In this paper ontology mapping approaches and methods in the e-learning are analyzed and discussed. We argue that to ensure effective reuse, mappings between learner profile ontologies, pedagogical context ontologies and domain ontologies should be done. We motivate the need both from performing static and dynamic ontology mapping. As there are several user profile modeling specifications (IMS LIP, PAPI Learner, eduPerson, Dolog LP, and others), specifications of mappings between them are needed. Static mappings can be performed between learning content domain ontologies, but usability of various resources for related concepts usually depends from concrete pedagogical goal and learning paths in different systems. We define concepts learning distance, and understandability as metrics for evaluating possibilities of usage of LO for learning concepts in different LMS or courses, specify and discuss some of it essential properties, as non-negativity, identity correctness. The paper presents a dynamic ontology mapping method in e-learning, using this metrics. As selection of learning material is in some degree subjective action, we present a fuzzy variant of this metric and use it to perform fuzzy mappings. We show the strengths and drawbacks of defined metrics and mappings, using examples, and discuss it applicability in practical e-learning systems.

Keywords: ontology mapping, E-learning, learning distance, fuzzy ontology mapping 1. Introduction
Ontology mapping (alignment) [3] is a process of discovering mappings between two ontologies through the application of a matching algorithm. Mapping elements are 5-tuple (IDij , n1i, n2j , Ri, Si) i = 1, ...,N1; j = 1, ...,N2; where IDij is a unique identifier (name) of the given mapping element; n1i is the i-th node of the first graph, N1 is the number of nodes in the first graph; n2j is the j-th node of the second graph, N2 is the number of nodes in the second graph; Ri specifies a similarity relation of the given nodes, and Si is a degree of confidence of given mapping. An alignment is a set of mapping elements. If some Si are less than 1, the alignment is called fuzzy. Ontology mapping is difficult and error-prone because of natural language ambiguity (synonymy, homonymy, etc.), and tack-dependence of knowledge conceptualization and representation. Learning materials are intended for some type of learners to achieve concrete learning goals in teaching clearly selected domain concepts (it meaning, relations to other concepts, etc.). To support reuse of learning resources between ontologically described e-Learning courses, differences in terminology in the description of metadata in different e-learning systems should be overcome. That is why mappings between learning profile ontologies, learning context ontologies, as well as between learning domain ontologies are needed.

September Issue

Page 28 of 105

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 3, No.3, 2011 There are several learning profile description standards, and definition of mappings between them is easy to be done. The pedagogical context can be represented by means of extra relations in domain ontologies, but it usually is specified by using of pedagogical ontologies (as learning path ontology, learning design ontology, learning goal ontology, learning context ontology). To support reuse of learning resources between ontologically described e-Learning courses, we will represent ontology mapping by taking into account different viewpoints of various users (teachers or learners) according to relatedness of concepts, defined or used in various courses or resources and different terminologies of various users, groups or even communities. Such ontology mapping we will call fuzzy learning goal driven mapping, as mapping is initiated in the process of searching of learning material for concrete learning goal. Our mapping is based on the new class similarity metric, called learning distance, or fuzzy learning distance enabling to take account both to the differences in learning paths, heterogeneity and consensus in collaborative learning environments. The partial differences in learning paths during the mapping or in concept definitions are evaluated by various viewpoints of various users and are presented as multi-viewpoint mappings between similar concepts in different courses.

2. Mapping learner profile ontologies

As various e-learning systems may use various learning profile representation standards, the mappings between every possible part of standards should be developed to facilitate reuse of appropriate for concrete learner resources between different systems. There are 6 main Learner Models[5]: PAPI Learner, IMS LIP, eduPerson, Dolog LP, FOAF, and Extended FOAF as a Learner Model[5]. PAPI Learner: PAPI presents learners information in six categories: Relations information, including learners relationships with other persons (teacher info, classmate, etc) , personal information, including general information about the student ( name, family, address, nationality, etc), preference information ( learning styles, preferred languages, authors, etc.), security information (public, private keys, passwords), performance information (for learning speed, assessment , grades, and certification), and portfolio information describing the learners projects and works, used for accessing their previous experience. PAPI ontology can be found in http://www.dcs.bbk.ac.uk/~kevin/selene/rdf/papi, but it is very schematic, and for practical usage extension is needed. IMS LIP is a part if IMS standard and was developed to model the lifelong records of learners achievement and to transfer their records between institutions and include information, focusing on the learners history and learning experience. Learners information in LIP is structured in eleven categories. The most important of them are general data about the learner, e.g. name, e-mail, information about learners qualification goals, certification, licenses (QCL), certifications, activities, interests, competencies, language capabilities, learning preferences. IMS LIP Ontology can be found in http://www.dcs.bbk.ac.uk/~kevin/selene/rdf/ims-lip.rdf, but it is very schematic, and for practical usage extension is needed. eduPerson is designed to facilitate communication between higher education institutions, ( mostly to move information about people between US universities). It includes detailed data about persons and organizations they are a member of, and may be used for exchanging data between other standards. It doesnt manage information about learners qualification goals, certification, licenses, activities, interests, competencies. Dolog LP uses RDF and learner ontologies, representing combined from PAPI and IMS LIP model to enable semantically enhanced learning systems to provide personalization services. It uses five main classes to describe the learner: personal info, preferences (proficiency, language, etc.), goal, interests, and study performance. FOAF (friend of-a friend) is an RDF vocabulary that provides a set of properties and classes to describe people, documents and organizations. It was developed for building communities and social groupings and is widely used in web 2.0 applications. FOAF distinguishes five categories for describing a person: FOAF Basics (name, e-mail, and images), personal Information (publications, weblog, points to people this person knows, and interests), Online Accounts, Projects and Groups, Documents and Images. It does not contain pedagogically-related information (as learning styles). In

September Issue

Page 29 of 105

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 3, No.3, 2011 [5] a taxonomy of the possible features that can describe a learner based on the analysis of the general structures of mentioned above learner models is presented. It may be used to extend FOAF (as one of the most frequently used ontologies) for building comprehensive learning profile ontology, incorporating information from various standards. Such ontology, or presented in [3] (TANGRAMs User Model Ontology) learner profile ontology may be used as reference ontologies to facilitate mappings between different user profiles in e-learning systems. LeMont is another IMS LIP based extension (by adding psychological learner preferences as cognitive or affective) There are a number of ontologies that were developed to extend the FOAF vocabulary in order to model the interests, hobbies, and preferences of a person ( for example Skill ontology, The FOAFCorp, extending FOAF with detailed description of the structure and interconnections of corporate entities, Description of a Career ontology (DOAC), FOAFRealm, etc. User profile information is understandable for usual users, but in many cases, all mappings are difficult to specify in design time, or make automatically because of persistence of many various standards (or it specific combinations). That is who visual easy-to use mapping tool is needed for the user to perform some possible mappings between user profiles of the two courses, he working with (if needed). For this purpose, we have download and try the Snoggle ontology mapping tool. It is Javabased graphical, ontology mapper for OWL ontology alignment. It allows users to visualize ontologies and visually draw mappings from one to another on a graphical canvas. Then Snoggle turns these mappings into SWRL/RDF or SWRL/XML format for use in the reasoning process. Snoggle can be downloading from http://snoggle.semwebcentral.org/

3. Mapping pedagogical context ontologies

As various e-learning systems may use various pedagogical models (and it ontological representation standards), the mappings between every existing ontologies, conceptualizing pedagogically-related knowledge should be developed to facilitate reuse of appropriate for particular pedagogical model resources between different systems. The LO structure and presentation, learning prerequisites, competencies and goals presentation are the main pedagogical context elements. An explicit definition of a LO structure can be useful in many cases if we need of special resource types (presentation, map, video, etc) or if we will reuse parts of a LO (for example to make adaptive content). Explicit definition of LO structure by using ontology make possible (semi)automatic LO annotation, division in small parts (as paragraphs, images, or location of some definitions) and decomposition oft LOs into ontology-defined components or recombination of some components into a new LOs. The Abstract Learning Object Content Model (ALOCoM) based ontology, which defines concepts that represent different types of content units and specifies their structure is used for this purpose. Such ontologies are relatively small and mapping them manually by means of a tool, having graphical interface (as Snoggle) is easy. [4] presents hierarchical structure of didactics-related ontologies to provide for the creators of courses improvements as reuse, rapid development and increased learning. An ontology that formalizes description of learning resources according to the IEEE Standard for Learning Object Metadata also has developed and tested [6]. This ontology adds the conceptual structure of learning object metadata and implements existing relations among them, in order to emphasize their pedagogical meaning, optimize learning processes helping students, professors, and designers of learning material, in automatizing processes of searching, acquiring, and using of learning materials. [7] Presents ontology for designing competency-based learning and knowledge management applications. It competency ontology is central to it instructional engineering methodology and is important before, during, and after LOs delivery. Before delivery, competencies ontology serves to guide the design or adaptation of a learning environment. During delivery, it guides the action of facilitators for learners assistance. After delivery, it helps assess the learning results and the evaluation of the quality of a learning environment. [8] presents an ontology to represent the semantics of the IMS Learning Design (IMS LD) specification. It facilitates the semantic description of the conceptual model as well as the definition of formal axioms related to both information and behavioural models in e-learning.

September Issue

Page 30 of 105

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 3, No.3, 2011 The Learning Goal Ontology for Collaborative Learning (http://www.ei.sanken.osakau.ac.jp/~ina/LGOntology/New/lgOntology.html) presents three kinds of goals: one common goal and two kinds of each member's personal goals. In this section, we distinguish among the three goals and identify the goals with justification based on learning theories. We have found a few more projects using pedagogical ontologies, every having it own standard. As by now there are only experimental projects in this area, and because of the specificity of pedagogical knowledge, only manual mapping of different such ontologies (by means of a tool, having graphical interface) will be appropriate in the process of it simultaneous usage. Some works are done in this direction: the OWL/XML code, proposed in [9], for example, is intended to make partial mapping between LOCO-Cite ontology and both the LOCO (Learning Object Context Ontologies) and ALOCoM Content Structure ontologies, using equivalentTo relation and hasLearningObject property. The comparison of the LOCO and LOCO-Cite ontologies to the LAMA ontology is the base for other analogical mapping. In some cases, user may use graphical mapping tool to perform some mappings.

4. Mapping domain ontologies

Domain ontologies in e-learning systems may be structured, following only semantic relations between concepts, or also taking into account the sequence of introducing the concepts in the presented course. For example, the concepts double and real as data types are relatively independent, and one course may define first double, and then real, and the second in the reverse order. Course domain ontologies may contain information about this order, or not. If not, we will say it pure domain ontology. If it contains information about the learning order, it is combined by the learning path ontology. It is clear, that for facilitating resource reuse, we will map only ontologies, describing closely related domains (in other cases LO reuse is not possible). Presented in this paper mapping method is not intended for direct initial mapping of domain ontologies. It supposed that many lexical and syntactic mappings between domain ontologies are already done. One of the main goals of our method is to facilitate LO reuse by estabilishing mapping of some LOs from one course to the ontological concepts of the other course (advising it as appropriate for learning or understanding this concept). In many cases, some ontological concepts of the two course ontologies are already mapped, but mapping may be uncertain (for example, if it is the result of automatic mapping process). In this case mapping of the same LO to the concepts of the two ontologies will increase the certainty of the previous syntactic mapping between these concepts. In other cases, when several resources, annotated as a resources for learning a concept from one e-learning course are mapped to the unmapped to it concept in the second course ontology, the mapping between the two concepts of different ontologies should be proposed on the base of common resources, used for learning these concepts. Using some learning objects for teaching related concepts in different systems make mappings between it ontological concepts in corresponding domain ontologies more certain.

4.1. Learning distance and understandability metrics

Suppose, that we wish to estimate possibilities to use the LO L, included in course K1 annotated as resource for teaching concept CL using it domain ontology O1 for learning a concept in course K2. To do this, we should estimate the possibilities of LO L to explain (in understandable way) the concept in course K2. We will first divide the domain terminology used in learning resource L in two main sets: RK1CL- related in some way to CL, and NRK1 CL, including all domain terms, dont related to CL. Elements of RK1 CL we divide in 3 subsets: Concepts, explaining CL, Concepts, explained by CL, and properties of CL Fig.1 We will use the following denotations: N number of all domain terminology in L, annotated by it domain ontology Np - number of classes (annotated by using LOs course domain ontology), used in explaining the semantic of the concept CL in the LO Ns - number of classes (annotated by using LOs course domain ontology), explained by the concept CL in the LO

September Issue

Page 31 of 105

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 3, No.3, 2011 Npr - number of data type properties of the concept CL, used in the LO Nnr - number of classes in the LO, dont related to the concept CL Definition 1: The ratio of the number of all the domain terminology, not related to concept CL in the LO L and all domain terminology, related to it we will call the Learning distance LD of the L to the concept CL. LD = Nnr /( Np+ Ns+ Npr+1) (1) ( 1 is for the term of concept name)

Learning resource L (Learning object, LO), used in course K1 Concepts, p1 explaining CL p2 p3 p4 a1

Hierarchical relations Object properties Figure 1. Terminology relation analysis for concept CL

a4

Learned concept CL pr1 pr2

Concepts, explain- 2 ed by CL s1 s3 s2 a2 2 a3 2

This metric is well defined, because of the denominator is never null. It obviously has non-negative value, and is 0 when and only when resource is intended to describe only the concept and it relations to the other concepts. It correctly estimate what part of the resource terminology is used to explain the corresponding concept (as greater is the number of terms, dont related to CL, so greater will be the LD). In other words, increasing of the number of terms, dont related to CL leads to increase of the distance, and upper equation clearly show that dependence. When LD is calculated, using the resource from one course, and the concept from domain ontology of other course, previously mapped to the first course domain ontology, it can be used to estimate possibilities of usage of the corresponding LO from the first course for teaching the concept in the second course. After mapping between domain ontologies of the two courses (only if the mapping between CL and concept in the other course ontology is successful) if the calculated for mapped concept in second course ontology and our resource LD value is significantly significantly greater, this may lead to the conclusion that LO L is not so appropriate for learning the concept CL in the new course. Another possible problem for resource reuse in other systems is it understandability. This means, that some of the terms, well known to the students, learning first course, may be unknown for students, attending second course, because of possible differences in the learning paths. More formally, we will define understandability CUCL of the concept CL in the new course as the ratio of the related to it terms, persisting in the undergone learning path, and all mapped related to the concept terms. We will use following denotations: NpH - number of classes, used in explaining the semantic of the concept CL in the LO, related to it by means of hierarchical relationships Npnh - number of classes, used in explaining the semantic of the concept CL in the LO, related to it by means of nonhierarchical relationships, NpmH - number of classes, used in explaining the semantic of the concept CL in the LO, related to it by means of hierarchical relationships, and mapped to at least one term from the CL mapped concept learning path in the second course ontology (concepts, known for the second course learner)

September Issue

Page 32 of 105

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 3, No.3, 2011 Npm - number of classes, used in explaining the semantic of the concept CL in the LO, related to it by means of nonhierarchical relationships, and mapped to at least one term from the CL mapped concept learning path in the second course ontology (concepts, known for the second course learner) Nsm - number of classes, explained by the concept CL in the LO, and mapped to at least one term from the CL mapped concept learning path in the second course ontology (concepts, known for the second course learner) Nprm - number of data type properties of the concept CL, used in the LO, and mapped to at least one term from the CL mapped concept learning path in the second course ontology (properties, known for the second course learner) Then we will calculate the numerical estimation of possible understandability of the concept CL, explained in LO L of the first course, if we use this resource L in the second course by mean of the formula: CUCL = (NpmH + * Npm +* Nsm+ Nprm)/ (NpH + * Npnh +* Ns+ Npr); Where and are coefficients, that recommended values are between 0 and 1, and represents relative significance of different types of relationships in explicit representation of concept semantic. This metric is correct because of it value is between 0 and 1, and is greater when greater number of related to CL concepts are mapped to the belonging to the corresponding of the CL concept learning path things in the second course ontology. The metric CUCL may be used for estimating domain-related understandability of resources (or it smaller parts), developed in the first course, before reusing them in the second course. Coefficients for understandability level of every concept may be also used.

4.2. Dynamic multistage domain ontology mapping method for e-learning

Mapping domain ontologies is difficult and in many cases error intensive tack because of natural language ambiguity and possible contextual differences between obviously related terms. In the elearning domain we should achieve higher precision and recall in mapping because of clear conceptual representation of the knowledge and presence of qualified textual resources, providing clear and unambiguous learned concept description. We present dynamic multistage ontology mapping method, which use textual learning resources and domain thesauruses (Fig. 2).

First e-learning course

E-learning repository 1 Thesaurus1

Second e-learning course

Thesaurus 2 E-learning repository 2

Text analysis module 1

Ontology mapping

Domain ontology 1

Domain ontology 2

Text analysis module 2

Ontology mapping

module 1

module 2

User interface 1

User interface 2

Figure 2. Conceptual schema of the mapping method On the first, initial mapping stage, lexical and syntactic (manual or automatic) mapping between domain ontologies of the two courses, between which we wish to establish resource reuse. Mappings are bidirectional, and with high level of uncertainty (higher if they are result of automatic mapping, or, lower if manual mapping by domain expert is performed). On this stage, the external domain thesauruses or ontologies may be used, but learning resources doesnt analyze and use. Every of both

September Issue

Page 33 of 105

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 3, No.3, 2011 systems may perform it own mappings, and may use some information, related to mappings, made by the other one. The aim of subsequent mapping stages is to check or increase certainty level of performed mappings (mark them as more, or less certain) or, find new possible mappings, using mutual learning resources. It is some type of structural, tree-based mapping. We first compare relations and related to the mapped objects in the two ontologies terms. If relations are the same and related terms are mapped, we marc considered mapping as more certain, but this is very rare situation: usually there are differences, and the main problem is to estimate the severity of this differences. To do this, we use LOs. For represented in the domain ontology concept, which mapping certainty we are checking, we search textual LOs, where it is defined (by using typical for definitions textual patterns), and make terminology relation analysis for this concept in chosen resources as shown in Fig. 1. Then we mark it explaining terms in the course domain ontology (consider them as the most important) and check it mappings to the terms of the other ontology. If they are mapped, and it corresponding concepts are related to the mapped concept in the ontology 2 by the same relations, we mark the considered mapping as certain. If some of explaining terms are not mapped, or mapped by uncertain mapping, we try to map them in the same way or make the same terminological analysis of explained terms of the corresponding concept in the second course, choose groups of explained terms and try to map them to explaining terms in the first ontology in the same way. If some of them are mapped to related to analyzed term in ontology 1, by certain mapping, we increase the certainty of this mapping, proportionally to the number of mapped terms, used in it explanation. If no one is mapped, we propose a low-security mapping between explained by the explained by most terms concept from second, and analyzed concept in the first ontology. The mapping direction is from first to second ontology. This is important to determine the direction of resource reuse. The described procedure is made one for every chosen concept in the second stage. Every time, when we make terminology relation analysis in textual resource for concept, for which we search the learning materials for reuse, we perform analogical mapping procedure (stage), related to the corresponding concept. We also propose directed low level certainty mapping (or make changes in the certainty of existing mappings) every time we find appropriate for reuse resource from one to another course. We will describe the mechanism of this type mapping later.

4.3. Dynamic domain ontology mapping and problems of resource reusing

We will specify and discuss the methodology of resource reuse between ontology described learning courses in related domains, using previously defined metrics and ontology mapping method, and discuss some experimental results, obtained in the process of the method testing. Our resource reusing methodology is as follow: 1. For resource reuse, we search other thematically-related, and described, using closely related metadata standards course. In the testing example, we use course in programming in C++ for reusing some resources, explaining basical programming concepts (for recalling) in the process of learning programming environments, using C++. 2. Make needed mappings between profile or pedagogical ontologies (if such ontologies are used) 3. Make possible lexical or syntactic mappings between domain ontologies, using available resources (domain lexicons, thesauruses, web terminological sources) 4. When user send the query for resource, and system cant find appropriate one in the curse, it will search in the external course. It is performing following: a. Analyze all resources, related to concepts, mapped to the concept, for which learning the resource is needed (if some mapping exists). If no mapping exists, the syntactic textual search in the external course will be performed. b. Calculate the Learning distance LD and understandability CUCL metrics for given concept in the chosen resources c. Divide resources to small parts to maximize it LD according to the interested concept d. Choose the best strategy of sorting resources according to the best combination of Learning distance LD and understandability CUCL e. Recommend the first few resources for reuse 5. Choose the best of the recommended resources according to the user profile and learning goal

September Issue

Page 34 of 105

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 3, No.3, 2011 Make needed changes in mapping between domain ontologies after choosing new resources for reuse. For experimental checking the valuability and usability of defined metrics, we have performed the following experiment: We have prepared in advance learning materials for teaching object-oriented programming, using C++, and for teaching programming environments, using C++, developed independently from each other and described by domain ontologies (included learning paths). When learning programming environments, in some cases, learners needed to remember meaning of basical concepts, or C++ syntaxes. This may be done by finding appropriate materials in the first, object-oriented programming course and add them as LOs to the second course. We examine and make some statistics of materials from the first course, used by students during performing two different tacks, in the second cource: learning for test, and doing concrete practical tack. Results are shown in table 1 and chart in figure 3. Averages are done on monitoring of 20 concepts. Table 1. LD and understandability evaluation in two learning tacks Average Average LD after LD after Average CU decom- ontology after decompoposition extension sition
5,41 0,99 0,39

6.

Average Average LD CU Learning for test Doing one practical tack

8,32 0,22

Average Average CU, number of according LR. for to learners learner

0,92 2,7

5,27

0,17

2,19

0,57

0,14

0,97

0,5

9 8 7 6 5 4 3 2 1 0

Learning for test Doing one practical tack

Av erage LD

Av erage CU

Av erage LD af ter decomposition

Av erage LD af ter ontology extension

Av erage CU af ter decomposition and ontology extension

Av erage CU, Av erage numaccording to ber of LR. f or learners learner

Figure 3. LD and understandability evaluation in two learning tacks graphical view On the base of these data, we make following conclusions: Average LD is high and average CU is very low. This leads to the conclusion, that resources in the external course contain much concepts, doesnt related to the needed ones, and concept explanation in these resources is not use known for the students terminology, according to the course ontology (in other words, external resources are unintelligible for the students). That statistics and very small number of initial mappings between two courses domain ontologies are because of the significant difference in the two domains terminology, according to it corresponding ontologies. Nevertheless, students have used more than 40 information sources from another course; level of real understanding is very high (witch contradict to the low values of CU). This means that many of the concepts in the first course are of some level known to the learners as a result of previous learning, second course use them implicitly as prerequisites, but not presented them in it domain ontology. We have performed ontology merging between the two course ontologies and then repeat the experiment using the resulting ontology as programming environments course ontology. This leads to mapping of almost all concepts of the first course ontology to the concepts of programming environments and significant change of

September Issue

Page 35 of 105

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 3, No.3, 2011 values of the two defined previously metrics. Nevertheless, the difference between values of the average LD and CU in the two tacks is significant. This means that in different tacks, the values of our metrics is of different importance. When, for example, the main aim of the learners is to understand and learn as many, as possible for many concepts, it is important that it and CU have larger value but LD may also be larger (learner prefer large resources, using several concepts simultaneously), but when the concrete property is of the main interest, LD metric is important to be lower, as user prefer small resources, closely related to the entity of interest. In these cases, resource decomposition is needed. It leads to increasing the understandability of resources and make them more usable for performing concrete tack. In other words, needed metrics values are learning goal-dependent.

5. Conclusions
This paper explores the problem of e-learning resource reuse in ontologically-based systems in the context of ontology mapping. It proposes definition and some evaluation of metrics for facilitating resource reuse, based on conceptual similarity and understandability of resources, belonging in one system (course) for learners of the other ones. It also presents multistage domain ontology mapping for e-learning method, performing lexico-syntactic and structural mapping, supported by thesauruses and textual sources (LOs). The structural stage usually requires mappings between other concepts and comparison of it learning paths to learning goal prerequisites. As selection of learning material is in some degree subjective action, and automatic ontology mapping is error intensive tack, we present a fuzzy variant of this metric and use it to perform fuzzy mappings. The understandability of external LOs is very important for it successful usage by learners. Our experimental evaluation leads to the conclusion, that automatical evaluation of understandability of external resources in supporting LOs reusing tack is correct only if ontological representation of all the learner knowledge (or all passing by them courses) is presented and used by the e-learning system.

6. References
[1] Barros, B. et. Al., Applications of a collaborative learning ontology, Micai 2002: Advances in Artificial Intelligence: Second Mexican International Conference on Artificial Intelligence, Mrida, Yucatn, Mexico, pp. 301-310, 2002. [2] Jelena Jovanovi et al, ANGRAM for Personalized Learning Using the Semantic Web Technologies , ournal of Emerging Technologies in Web Intelligence, Vol 1, No 1, pp. 6-21, 2009. [3] Kalfoglou, Y., Schorlemmer, M., Ontology mapping: the state of the art, The Knowledge Engineering Review, 18(1), pp. 131, 2003. [4] Mencke S., R. R. Dumke, A Hierarchy of Ontologies for Didactics-Enhanced E-learning, Conference ICL2007 September 26 - 28, 2007. [5] Ounnas, A., et al., Towards a Semantic Modeling of Learners for Social Networks, in Proceedings of SW-EL workshop at AH2006, pp. 102-108., 2006. [6] Romero, L. et al., An ontology for semantic definition of learning objects, Information Systems and Technologies (CISTI), 5th Iberian Conference on Issue Date: 16-19 June 2010 pp. 1 7, 2010 [7] Paquette, G., An Ontology and a Software Framework for Competency Modeling and Management, Educational Technology & Society, 10 (3), pp. 1-21, 2007. [8] Amorim, R. R., et al., A Learning Design Ontology based on the IMS Specification, Educational Technology & Society, 9 (1), pp. 38-57, 2006. [9] Knight C., et al, "Ontologies for context-dependent reuse of learning designs and learning content, in Proc Second Annual Conference, 2005.

September Issue

Page 36 of 105

ISSN 2229 5216

International Journal of Advances in Science and Technology, Vol. 3, No.3, 2011

Author Profile
Dr. Tatyana Ivanova Ivanova received his Ph.D degree from Technical University, Sofia, Bulgaria in 2009. A life member of John Atanasoff Union of Automation and Informatics and Union of Electronics, Electrical Engineering and Telecommunications (CEEC). She is currently working as a tutor in college of energy and electronics, department of the Technical Univercity of Sofia. She is having overall teaching experience of 21 years including high scools. Shes major research Interests are in Semantic web, Expert systems, Artificial Intelligence, database and Computer-Engineering.

Photo Goes here

September Issue

Page 37 of 105

ISSN 2229 5216