keynote taiwan
TRANSCRIPT
Prof. Dr. Arif ALTUNHacettepe University / Ankara-Turkey
Keynote Presentation at IETC 2012 Taiwan
Ontologies for Personalization: A new challenge for instructional designers
IETC 2012
Personalization
• Personalization is described as adapting learning experiences to different learners by analyzing individuals’ knowledge, skills and learning preferences (Devedzic, 2006).
• …tailors instructional materials for each learner’s constantly changing needs and skills (Sampson, Karagiannidis & Kinshuk, 2002).
Five types of personalization
1. Name based personalization2. Self-described personalization3. Segmented personalization4. Cognitive personalization5. Whole-person personalization
(Martinez, 2000).
Some of the Challenges for ID
• Paradigm shift: From “one design for one learner” to “many designs for one learner ”
• Better understanding the nature and the outcomes of the interaction between learners and content.
• Designing learning objects • Designing navigational paths• Monitoning and analyzing the learning progress• But, how should we proceed?
In order to make an e-learning environment personalized,
– Regular and constant data monitoring and analysis tools (Learning Analitics),
– Determining cognitive and non-cognitive personal characteristics accurately, (Learner characteristics)
– Learners’ interaction with –designed- medium: i.e., learning outcomes (Learning & Instruction)
– Tools to diagnose and/or guide learners with study or navigational paths (Ontology and Designing Navigational Paths).
What we need is
1. A learner model2. A learning object design model3. Ontolog(ies)4. Learning analytics
International Conference Cognition and Exploratory Learning in Digital Age CELDA 2010
www.ontolab.hacettepe.edu.tr/en
A Learner Model for Learning Object Based Personalized Learning Environments
• What will be modeled about learners?• How will it be modeled? And,• How the sustainability of the model would be
maintained?
Kaya & Altun, 2011
Neuropsychological Assessment
• Determining the strengths and weaknesses in one’s cognitive functions (such as, memory types, attention levels, language ability etc.)
• Paper-pencil tests vs computerized tests
Line Orientation Test
Enhanced Cued Recall Test
Test Environment
• ECRT– no correlation was observed between computerized and P&P tests (r= -.09; p>.05)
• Significant correlation was observed in LOT (r= .61; p<.05)
• ECRT– P&P test scores are higher than ( M= 46.07; SD= 2.127 ) computerized one (M = 40.12; SD= 5.099).
• LOT– P&P test scores are higher than (M= 22.76; SD = 4.314) computerized one (M= 19.58; SD= 4.933).
• ECRT and LOT: Time spent in P&P tests is much longer than the computerized one.
What do results imply?
• Sönmez, D., Altun, A. & Mazman, S. G. (2012). How Prior Knowledge and Colour Contrast Interfere Visual Search Processes in Novice Learners: An Eye Tracking Study. Under Review.
• The effect of persons’ prior knowledge and experiences on their visual search performances.
• A visual search task on identifying the phases of mitosis from a microscope view with two different background contrasts.
Visual Search
Low level prior knowledge High level prior knowledge Prior exposure
(n=10)No Prior exposure
(n=10)Prior exposure
(n=10)No Prior exposure
(n=10)
Blue (High Contrast)
Fix.Dur.M 1.46 1.29 2.79 2.81Sd .806 .764 1.27 1.94
T_FirstFix.M 7.04 7.52 4.66 3.98Sd 5.2 5.07 3.61 4.33
Yellow (Low
Contrast)
Fix.Dur.M .818 .946 1.28 .889Sd .728 .813 .852 .697
T_FirstFixM 4.93 3.52 3.84 5.18Sd 3.24 2.87 2.31 5.22
• Different STM spans (High - medium - low) undergraduate students in two different attention design types: (Focused-divided)
• Dependent variable : recall performance• Time spent in focused one is longer than in divided design• Recall performance is affected across modalities: Low STM <
High STM and Meed STM < high STM• Low STM group spent more time in the environment than the
High STM group
Short-term memory spans and attention design
• Different location memory groups• Dependent variable: Recall performance• Environment: 2-D vs 3-D environments.
Spatial Location Memory and Navigation Environment
Findings
• Overall, participants had higher recall scores in 2D.• Once controlled their location memory, however, results
indicate that higher LM group had higher recall scores in 2D, but did not change for low LM group.
• Male participants were advantageous over females in 3-D.
Dependent Variables: Recall and retention (free recall, heading recognition, and location memory)
Levels of Processing and Navigation design
Heading recognition task
Location memory task
• Left side navigation menu yielded better results in free recall, heading recognition, and location memory
• Deep level of processing yields better recall performances
• Memory performances are affected depending on the design of the given instruction (levels of processing).
Challenges
• More research is needed across age groups, gender, and in culturally different settings.
• How much time is needed?• How to differentiate the learning paths for
individuals and/or group of learners?
Learning Objects
Some definitions to start with…• A learning object is defined as “…any entity,
digital or non-digital, that may be used for learning, education or training” (IEEE Learning Technology Standards Committee, 2001).
• “...a Learning Object... [is] ‘any digital resource that can be reused to support learning” Wiley (2002).
Common Characteristics of LOs• All learning objects need to have an
instructional purpose to be re-used within different instructional settings.
• Each LO should appropriately support learning through the possible inclusion of educational objectives, content, resources, and assessment.
Common Metaphors• Lego (i.e., Hodgins & Conner, 2000)• Learning Atom & Learning Crystal (Wiley, 2001)• Luggage (Dawnes, 2002)
Fundemental Questions for IDs• How to store each learning object so that they
can further become accessible through different digital learning and/or content management systems or different delivery modes
• What should be the size of the learning object (granuality)
• How can the context be modeled?
Learning Space Model Aşkar & Altun (2010)
• Proposes a separation of learning expectations as concepts and skills based on their ontological relations in a specific domain;
Ontology based representation of A Learning Object
Adjustable Relation
Concept Space Skill Space
1
Raw Content
LC
2 3 4 n
Adjusted Weight via Intelligent Bot
Content
LC
2 3 44 n1
Calculated (or pre-defined) Relation via Intelligent Bot
Content
LC
2 3 44 n
Calculated (pre-defined) Relation via Intelligent Bot
1
Ontology-based Learning Space
ConceptsLearning Space (LS)
Learning Container (LC)
Learning Objects (LO)
Assets
Adjusted Weight
Skills
Representation of skills and concepts in ontology space
21
22
Representation of skills and concepts in ontology space
Challenges
• Reusable,• With reasonable granuality,• Capable of handling learning contexts, • Interoprable, and• LO development tools (designed with an
instructionally sound design approach) are needed.
ONTOLOGY
An ontology is …
• an explicit specification of a conceptualization (Gruber, 1995) or a model (Musen, 1998), which is used for structuring and modeling of a particular domain that is shared by a group of people in an organization (O’Leary, 1998).
• Domain ontologies provide explicit and formal descriptions of concepts in a domain of discourse, their properties, relationships among concepts and axioms (Guarino, 1995)
Semantic Web– Well defined meanings (semantics)– Common and shared standards and technologies
Tim Berners-Lee
The challenge is…
• By using the capabilities of semantic web, World Wide Web led the interchange of information about data (e.i., metadata) as well as documents.
• Such capabilities also indicated a new kind of challenge for instructional designers to design a common framework that allows content to be shared and reused within and across applications.
Stage 1: Identifying the conceptsStage 2: Determining class and class hierarchies Stage 3: Determining the attributes within classes and their relationshipsStage 4: Determining instancesStage 5: Setting up axioms / rules
(adapted from McGuiness, 1999)
Ontology as a Design & Development Process
PoleONTO: Modeling the K-12 curricula by using ontology
PoleONTO Personalized Ontological
Learning Environments
Expectation
Expectation 2
Expectation ..n
Concept
C1
C n
C2
Skill
S1
S n
S2
Expectation
• CogSkillNet is an ontology of skills exists in the curriculum of K-12 education.
• In POLEonto context, skill is defined as the interaction and any processes between persons and concepts. For example, the concept of “square” is envisioned in one’s mind; yet, they can define it, they can extend square into some other thing (i.e., a table or a flower-stand), which is creative thinking. The square can be manipulated to approach a problem by using its types and functions, which requires problem solving.
• Expectations in K-12 curricula
• Cognitive action verbs in curricula– Put, show, etc.– Summarize, generalize,
etc.– Critical thinking, problem
solving, etc.
Identifying the concepts class and class hierarchies attributes within classes and
their relationships Determining instances Setting up axioms / rules
Identifying the concepts class and class hierarchies attributes within classes and
their relationships Determining instances Setting up axioms / rules
• Y: is an instance of • X: is a class of• C: is a superClass of• A: is a subClass of • K: is a process_component of• T: has process_component of
Skills Relation Skills
Integrated Skill X Analyze
Analyze Y Integrated Skill
Analyze T Determine RelationshipDetermine
relationshipK Analyze
Basic Skill C Encapsulated SkillEncapsulated Skill A Basic Skill
Identifying the concepts class and class hierarchies attributes within classes and
their relationships Determining instances Setting up axioms / rules
Identifying the concepts class and class hierarchies attributes within classes and
their relationships Determining instances Setting up axioms / rules
• Each act can be acted upon.• Each action can include sub-actions.• All actions can call others while being executed.• All actions start with an input and produces an
output.• An Output can be an input for another action.• Inputs and outputs can be null, single or multiple.
Identifying the concepts class and class hierarchies attributes within classes and
their relationships Determining instances Setting up axioms / rules
Taxonomic View of CogSkillNet
From taxonomy to ontology
Some Screenshots
Design and Application of Apothegm Ontology
• 90 apothegmes were selected • 281 concepts with 113 action verbs• Relations:
– hasMeaning (isMeaningOf),– hasComponent (isComponentOf),– hasMeaningValue (isMeaningValueOf)
Visualizing the ontology
• A web based navigation tool is designed• Apothegmes were presented on screen, users
navigate by selecting an apothegm and reaches its components, meaning, and type.
• In addition, users are provided an interface in order to add new statements and relations to the ontology.
Apothemes.owl
Semantic web tools
UI
Ontology
Visual Representation
Compenents when selected an apothegm
To conclude…
• Personalization can be a valuable tool to facilitate lifelong learning with just-in-time and on-the-job training, as well.
• Different frameworks and learner (and group) characteristics will drive the method of personalization
• Personalization can be expensive and time-consuming if properly developed and maintained
Last but not the least…Davie & Inskip (1992) once emphasized
“good instructional design is more important than the specific technology”
and, Ana Donaldson puts it well
“ online courses are demanding further considerations”
…thus, we need to “know our learners well”
Thank you for your patience…Hacettepe University , Computer Education and Instructional Technologies
Thank you...
For the list of references, see http://www.ontolab.hacettepe.edu.tr and/orhttp://www.ontolab.hacettepe.edu.tr/en