Bayesian Network Student Model for Adapting Learning Activity Tasks in Adaptive Course Generation...

1
Bayesian Network Student Model for Adapting Learning Activity Tasks in Adaptive Course Generation System Introduction Adaptive educational hypermedia system (AEHS) aims to develop the course that can adapt to users. One important components of it is learner model. Learner model represents individual learner’s information such as knowledge, background, learning goals, learner’s preference, etc… that useful for adaptation. In this paper, we represent the learner modeling component of ACGS, and how to develop Bayesian Network (BN) learner model to manage overlay knowledge model and adapt learning activities based on task model. In addition, we describe an implementation of this model for computer science domain, a database course subject: “How to design relationship database?”. Course domain model of ACGS. Learning activity adaptation Adaptation is process to select activity tasks for each learner based on learner’s model. The learner with different knowledge level needs to do different tasks in order to finish learning goal. This is composed by several tasks which include abstract tasks and activities task. Viet Anh Nguyen a , Viet Ha Nguyen a , Si Dam Ho a , Hitoshi Sasaki a College oF Technology, Vietnam National University Hanoi, Vietnam b Faculty oF Engineering, Takushoku University, Japan Background This section describes several theoretical backgrounds which involved our research. What can be adapted? •macro-adaptive selecting a few components that define the general guidelines for the e-Learning process, such as learning objectives or levels of detail and mainly based on learner model. •aptitude-treatment proposing different types of instructions and/or different types of media for different students. •micro-adaptive, diagnosing the student’s specific learning needs during instruction, providing instructional prescriptions for these needs and monitoring the learning behavior of the student while running specific tasks and adapting the instructional design afterwards, based on quantitative information. The domain model Domain model is an object model of problem domain. In AEHS, domain model is set of elements about educational domain; each element is domain object class and the relationship between them. Domain model decompose knowledge of the subject into fragments such as: topic, sub-topic, atomic concepts. Depending on the domain, designer strategies, there many kinds of domain model structure: vector model, network model, and ontology, etc. The overlay knowledge model The overlay model is one that supposes the student’s knowledge to be a subset of the system’s knowledge of the subject. As the student learns, the subset grows, and the modeler’s job is to keep trace of the subset. This model assumes that the student will not learn anything that the expert does not know. The principle of the learner’s overlay model is that for each domain model concepts, individual user knowledge model store data that represent values which is an estimation of the user knowledge level of this concept. The task model A task statement refers to a set of coherent activities that are performed to achieve a goal in a given domain. Task models are documentation structures that are used for: i) documenting the result of a task design of proposed Conclusion The main contribution of this paper is a model to manage student model based on learner overlay knowledge model. As a result, model gathers information about learner current state of knowledge and modeling learner as unreliable source of concepts. Improving our previous work, adaptation process extends to adapt learning activity task based on task model in order to adapt for know-how and learner’s learning goals. For this, prerequisite relationship among activity task as taking into account for selecting learning material process. Finally, we developed ACGS architecture for generating adaptive course. For more detail, please contact with: Viet Anh Nguyen Email: [email protected] Web: http://www.coltech.vnu.edu.vn/~anhnv Selected References Nguyen Viet Anh, Nguyen Viet Ha, Ho Si Dam (2008). " Contructing a Bayesian Belief Network to Generate learning path in adaptive hypermedia system".Journal of Computer Science and Cybermetics Vol 1(24), 2008, p. 12-19. Viet Anh Nguyen , Si Dam Ho (2006), "Applying Weighted Learning Object to Build Adaptive Course in E-learning", Learning by Effective Utilization of Technologies: Facilitating Intercultural Understanding, Frontiers in Artificial Intelligence and Applications, Volume 151, p 647-648, Beijing, China. Viet Anh Nguyen, Si Dam Ho (2006)."ACGS: Adaptive Course Generation System- An efficient approach to build E-learning course".Proceeding of 6th IEEE International Conference on Computers and Information Tecnology, 2006, p 259- 265,Seoul, Korea. Adaptive Course Generation System (ACGS) Course domain model includes several topics which include two objects are concepts and learning tasks as depicted in figure 2. In order to acquire a concept, learner need to work several related learning tasks. Figure 4. Adaptation engine of Adaptive Course Generation System (ACGS) Captions to be set in Times or Times New Roman or equivalent, italic, between 18 and 24 points. Right aligned if it refers to a figure on its right. Caption starts right at the top edge of the picture (graph or photo). LearnerModule Visualization M odule Adaptation engine Learning O bject D atabase Learnermodel/ Learnerprofile Adaptation M odule Figure 1. Adaptive course generation architecture ACGS (NGUYEN & Dam, 2006) includes three modules: Learner Module (LM), Visualization Module (VM) and Adaptation Module (AM) as depicts in figure 1. Learner Module designed to get learner’s demand such as learning goals, preferences, etc. and to evaluate learner’s knowledge. Visualization Module takes adaptive course outlines for displaying them as annotated hypertext links in the website to learner. Adaptation Module asks domain concepts from Learning Object Database which includes learning resources and learning-task database, which contains tasks with all possible combinations of levels of support and complexity as well as enough variability over other task features to allow for generalization and abstraction by the learner as well as asks learner’s knowledge, and learner’s learning goals to generate course structure. C ourse D om ain M odel Topic C oncepts Learning Tasks Topic Topic Figure 2. Course domain model of ACGS Otherwise in order to finish learning task, learner also needs to acquire some concepts corresponding. The course domain is represented as a directed acyclic graph (DAG) with several nodes and vertex connects between them. Node depicts an atomic concept while vertex depicts prerequisite relationship between the concepts. Task1 D esign Entities R elationship D iagram Task 3 N orm alize Tables Task 2 Transform E R D to Tables Physic Identify Tabbles Identify R elationships am ong Entities Define C onstraints Transform to 1 st N orm al Form Transform to 2nd N orm al Form Transform to 3tr N orm al Form D efine Attributes ofEntity Identify E ntities H ighlight N ouns (things, people, organization) C heck nouns is com m on noun and single m eaning? Identify N ouns,num ber w hich type or characters of entities C heck attributes has atom ic values? D efine dom ain value of attributes C heck attributes can “answ er” question of problem s? C hoose one or m ore attributes thathas value unique identifies entity (C hoose key) H ighlightVerbs describe relationships betw een entities S hort description relationship link Identify R elationship C ardinality (H ighlight num bers) Identify relationship is is the participation oroptionally Task & Sub Task: Task/Techniques: Sub Task relation: Sub Task/Technique consequence : Transform entities to table nam es D efine field types,dom ain value Transform attributes to fields Identify prim ary key Bring key from “one”table to “m any”table (break one-to- m any) C reate new table w ith fields are prim ary key oftw o tables w hich have m any to m any relation (break m any to m any) D efine constraints for table level D efine constraints for Field level U nderline repeating attributes group the prim ary key forthe original relation is included in both ofthe new relations U nderline non- key attributes dependenton partofprim ary key attributes D ecom pose tw o tables w ith one forrepeating groups and one non-repeating groups separate outall the attributes thatare solely dependenton partofprim ary key attributes rem ove the attributes involved in the transitive dependency and putthem in a new relation U nderline non- key attributes dependenton partofnon-key attributes Task 4 D efine Q uery to R etrieve D efine queries to D efine D atabases (Ex: C reate table … ) D efine queries to m anipulation data (ex:Insert, U pdate, D elete… ) D efine queries to retrieve inform ation Figure 3. Partial learning activity task of How to design relationship database” course Bayesian Network learner model To develop BN learner model, we assign a set of variables to measure learner’s knowledge with three states: not acquired, in progress, acquired. p(not-acquired(C)) represents probability value of not acquired state for concept C, p(in-progress(C)) denotes probability value of in progress state for concept C, and p(acquired(C)) denotes probability value of acquired state for concept C. Takes Q uestionnaires C hoose learning goals Constructing Dom ain C oncepts U pdate Learner Profile C onstructing Learning path Selecting Learning path /R esource Selecting Learning R esource G etting R esource [no adaptation] Adaptive System Learner Adaptation process selects learning resources through phases: First of all, resources are evaluated and classified in one equivalence class according to class membership rules are selected base on learner profile and adaptation rules (NGUYEN & Dam, 2008b) which is a set of rules represented in first order logic. Secondly, according to adaptive navigation technique, one ore more techniques is selected such as hiding, annotation or direct guidance in order to input for visualization module to display the course to choose for current learner. Finally, student activities response will be updated in his/her profile which is basic for adaptation process in next run-time learning activities. Our experiments We design a course topic “How to design relationship database?” for third year student. In order to design database, first of all the student need to skim problem’s speciation and then participates four phrases: designing entities relationship diagram, transforming entities relationship diagram to tables physic, normalizing tables, and defining query to retrieve information. There are twenty six concept nodes in the course model. The task diagram is composed by twelve abstract tasks and twenty nice activity tasks. There are two kinds of activity tasks: consequent task and parallel task.

Transcript of Bayesian Network Student Model for Adapting Learning Activity Tasks in Adaptive Course Generation...

Page 1: Bayesian Network Student Model for Adapting Learning Activity Tasks in Adaptive Course Generation System Introduction Adaptive educational hypermedia system.

Bayesian Network Student Model for Adapting Learning Activity Tasks in Adaptive Course

Generation System

IntroductionAdaptive educational hypermedia system (AEHS) aims to develop the course that can adapt to users. One important components of it is learner model. Learner model represents individual learner’s information such as knowledge, background, learning goals, learner’s preference, etc… that useful for adaptation.

In this paper, we represent the learner modeling component of ACGS, and how to develop Bayesian Network (BN) learner model to manage overlay knowledge model and adapt learning activities based on task model. In addition, we describe an implementation of this model for computer science domain, a database course subject: “How to design relationship database?”.

Course domain model of ACGS.

Learning activity adaptationAdaptation is process to select activity tasks for each learner based on learner’s model. The learner with different knowledge level needs to do different tasks in order to finish learning goal. This is composed by several tasks which include abstract tasks and activities task.

Viet Anh Nguyena, Viet Ha Nguyena, Si Dam Hoa, Hitoshi SasakiaCollege oF Technology, Vietnam National University Hanoi, Vietnam

bFaculty oF Engineering, Takushoku University, Japan

BackgroundThis section describes several theoretical backgrounds which involved our research. What can be adapted?•macro-adaptive selecting a few components that define the general guidelines for the e-Learning process, such as learning objectives or levels of detail and mainly based on learner model.•aptitude-treatment proposing different types of instructions and/or different types of media for different students.•micro-adaptive, diagnosing the student’s specific learning needs during instruction, providing instructional prescriptions for these needs and monitoring the learning behavior of the student while running specific tasks and adapting the instructional design afterwards, based on quantitative information.

The domain modelDomain model is an object model of problem domain. In AEHS, domain model is set of elements about educational domain; each element is domain object class and the relationship between them. Domain model decompose knowledge of the subject into fragments such as: topic, sub-topic, atomic concepts.

Depending on the domain, designer strategies, there many kinds of domain model structure: vector model, network model, and ontology, etc.

The overlay knowledge model The overlay model is one that supposes the student’s knowledge to be a subset of the system’s knowledge of the subject. As the student learns, the subset grows, and the modeler’s job is to keep trace of the subset.

This model assumes that the student will not learn anything that the expert does not know. The principle of the learner’s overlay model is that for each domain model concepts, individual user knowledge model store data that represent values which is an estimation of the user knowledge level of this concept.

The task modelA task statement refers to a set of coherent activities that are performed to achieve a goal in a given domain. Task models are documentation structures that are used for: i) documenting the result of a task design of proposed activities, ii) supporting personnel selection, iii) identifying needs for training.

Bayesian networkA BN is a directed graph whose nodes represent the (discrete) uncertain variables of interest and whose edges are the causal or influential links between the variables.

Conclusion

The main contribution of this paper is a model to manage student model based on learner overlay knowledge model. As a result, model gathers information about learner current state of knowledge and modeling learner as unreliable source of concepts.

Improving our previous work, adaptation process extends to adapt learning activity task based on task model in order to adapt for know-how and learner’s learning goals. For this, prerequisite relationship among activity task as taking into account for selecting learning material process. Finally, we developed ACGS architecture for generating adaptive course.

For more detail, please contact with:Viet Anh NguyenEmail: [email protected]: http://www.coltech.vnu.edu.vn/~anhnv

Selected ReferencesNguyen Viet Anh, Nguyen Viet Ha, Ho Si Dam (2008). " Contructing a Bayesian Belief Network to Generate learning path in adaptive hypermedia system".Journal of Computer Science and Cybermetics Vol 1(24), 2008, p. 12-19.

Viet Anh Nguyen , Si Dam Ho (2006), "Applying Weighted Learning Object to Build Adaptive Course in E-learning", Learning by Effective Utilization of Technologies: Facilitating Intercultural Understanding, Frontiers in Artificial Intelligence and Applications, Volume 151, p  647-648, Beijing, China.

Viet Anh Nguyen, Si Dam Ho (2006)."ACGS: Adaptive Course Generation System- An efficient approach to build E-learning course".Proceeding of 6th IEEE International Conference  on Computers and Information Tecnology, 2006, p 259- 265,Seoul, Korea.

Adaptive Course Generation System (ACGS)

Course domain model includes several topics which include two objects are concepts and learning tasks as depicted in figure 2. In order to acquire a concept, learner need to work several related learning tasks.

Figure 4. Adaptation engine of Adaptive Course Generation System (ACGS)

Captions to be set in Times or Times New Roman or

equivalent, italic, between 18 and 24 points. Right aligned if

it refers to a figure on its right. Caption starts right at

the top edge of the picture (graph or photo).

Learner Module

Visualization Module

Adaptation engine

Learning ObjectDatabase

Learner model/ Learner profile

Adaptation Module

Figure 1. Adaptive course generation architecture

ACGS (NGUYEN & Dam, 2006) includes three modules: Learner Module (LM), Visualization Module (VM) and Adaptation Module (AM) as depicts in figure 1.

Learner Module designed to get learner’s demand such as learning goals, preferences, etc. and to evaluate learner’s knowledge.

Visualization Module takes adaptive course outlines for displaying them as annotated hypertext links in the website to learner.

Adaptation Module asks domain concepts from Learning Object Database which includes learning resources and learning-task database, which contains tasks with all possible combinations of levels of support and complexity as well as enough variability over other task features to allow for generalization and abstraction by the learner as well as asks learner’s knowledge, and learner’s learning goals to generate course structure.

Course Domain Model

Topic

ConceptsLearning Tasks

Topic Topic

Figure 2. Course domain model of ACGS

Otherwise in order to finish learning task, learner also needs to acquire some concepts corresponding. The course domain is represented as a directed acyclic graph (DAG) with several nodes and vertex connects between them. Node depicts an atomic concept while vertex depicts prerequisite relationship between the concepts.

Task1Design Entities Relationship

Diagram

Task 3Normalize Tables

Task 2Transform ERD to Tables

Physic

Identify Tabbles

Identify Relationships

among Entities

Define Constraints

Transform to 1st

Normal Form

Transform to 2nd Normal Form

Transform to 3tr

Normal Form

Define Attributes of Entity

Identify Entities

HighlightNouns (things,

people, organization)

Check nouns is common noun

and single meaning?

IdentifyNouns, number which type or characters of

entities

Check attributes has atomic

values?

Define domain value of

attributes

Check attributes can “answer” question of problems?

Choose one or more attributes that has value

unique identifies entity (Choose

key)

Highlight Verbs describe

relationships between entities

Short description

relationship link

Identify Relationship Cardinality (Highlight numbers)

Identify relationship is is the participation

or optionally

Task & Sub Task: Task/ Techniques: Sub Task relation: Sub Task/Technique consequence :

Transform entities to table

names

Define field types, domain

value

Transform attributes to

fields

Identify primary key

Bring key from “one” table to “many” table

(break one-to-many)

Create new table with fields are primary key

of two tables which have

many to many relation

(break many to many)

Define constraints for

table level

Define constraints for

Field level

Underline repeating

attributes group

the primary key for the original

relation is included in both

of the new relations

Underline non-key attributes dependent on part of primary key attributes

Decompose two tables with one for repeating

groups and one non-repeating

groups

separate out all the attributes that are solely dependent on part of primary key attributes

remove the attributes

involved in the transitive

dependency and put them in a new relation

Underline non-key attributes dependent on part of non-key

attributes

Task 4Define Query to Retrieve

Define queries to Define

Databases (Ex: Create table …)

Define queries to manipulation data (ex: Insert,

Update, Delete…)

Define queries to retrieve information

Figure 3. Partial learning activity task of “How to design relationship database” course

Bayesian Network learner modelTo develop BN learner model, we assign a set of variables to measure learner’s knowledge with three states: not acquired, in progress, acquired. p(not-acquired(C)) represents probability value of not acquired state for concept C, p(in-progress(C)) denotes probability value of in progress state for concept C, and p(acquired(C)) denotes probability value of acquired state for concept C. there is p(not-acquired(C)) + p(in-progress(C)) + p(acquired(C)) =1.

Takes QuestionnairesChoose

learning goals

Constructing Domain Concepts

Update Learner Profile

ConstructingLearning path

Selecting Learning path/ Resource

Selecting LearningResource

Getting Resource

[no a

dapta

tion]Adaptive

System

Learner

Adaptation process selects learning resources through phases:•First of all, resources are evaluated and classified in one equivalence class according to class membership rules are selected base on learner profile and adaptation rules (NGUYEN & Dam, 2008b) which is a set of rules represented in first order logic.•Secondly, according to adaptive navigation technique, one ore more techniques is selected such as hiding, annotation or direct guidance in order to input for visualization module to display the course to choose for current learner. •Finally, student activities response will be updated in his/her profile which is basic for adaptation process in next run-time learning activities.

Our experiments

We design a course topic “How to design relationship database?” for third year student. In order to design database, first of all the student need to skim problem’s speciation and then participates four phrases: designing entities relationship diagram, transforming entities relationship diagram to tables physic, normalizing tables, and defining query to retrieve information. There are twenty six concept nodes in the course model. The task diagram is composed by twelve abstract tasks and twenty nice activity tasks. There are two kinds of activity tasks: consequent task and parallel task.