Recommendation of Educational Resources in a Blended Learning

Environment

Diego Alessandro Pereira dos Santos

1,2 a

, Isabela Gasparini

and

José Palazzo Moreira de Oliveira

Instituto de Informática, PPGCC, Federal University of Rio Grande do Sul, Brazil

Federal Institute of Education, Science and Technology Sul-rio-grandense (IFSul), Sapiranga, RS, Brazil

Universidade do Estado de Santa Catarina: Joinville, Santa Catarina, Brazil

Keywords: Ontologies, Blended Learning, Recommender Systems, Context-Aware Systems.

Abstract: Blended learning environments are those that combine face-to-face instruction with computer-mediated

instruction and have gained space in the means of discussion about new educational methodologies. Several

benefits are observed in the use of this methodology, among them: an increase in academic performance and

students' social skills, an increase in teaching and learning flexibility, an increase in student satisfaction, à

decrease in dropout rates, and an increase in school retention. Recommender systems are useful in these

environments, providing the suggestion of content and activities personalized to users; here, we present a

model for recommending learning activities in a blended learning environment. To evaluate the model, SWRL

rules were used through the Pellet inference engine. The approach was evaluated through a case study that

represents the situation of a student in a blended learning environment, with several options of activities, in

which the choices may vary according to their general and academic profiles, in addition to their context. The

recommendation rules are executed, resulting in the activity suggestion for the student. Thus, it was verified

that the developed model fulfills the proposed objective of enriching the recommendation of learning

resources in a blended learning environment through the modeling of the learner's profile and of the

educational resources with context awareness through ontology.

1 INTRODUCTION

The educational environment has undergone several

digital changes in recent years, mainly due to cultural

evolution and increased access to technological

advances by the general population. Most students

use mobile devices in the classroom, even when these

devices do not comply with established norms. Often

this movement is an attitude of an initiative to search

for information, which shows an opportunity for

development in computing and education. In this

context, hybrid teaching has regained strength in the

spaces of discussion about new educational

methodologies. Blended learning has been studied

and disseminated for several years (Oliver, 2005)

(Graham, 2006). As early as 2003, the American

Society for Training and Development (ASTD)

https://orcid.org/0000-0002-9578-2601

https://orcid.org/0000-0002-8094-9261

https://orcid.org/0000-0002-9166-8801

already identified blended learning as one of the ten

most significant trends in the knowledge industry

(Rooney, 2003). In current work, researchers have

been investigating both the implementation and the

effects of blended learning in different educational

environments (Rafiola et al., 2020) (Vo et al., 2020)

(Anthony et al., 2020) (Bruggeman et al., 2021).

More recently, Agarwal (2021) dealt with some

trends in educational practices in the coming years,

pointing out blended learning as one of the most

promising practices in the near future. It shows that

recent events, such as the COVID-19 pandemic,

accelerated the process of change in several

educational practices that were already being thought

of, highlighting the role of hybridity in this

movement. Among the practical advantages of

blended learning, some stand out, according to the

Santos, D., Gasparini, I. and Moreira de Oliveira, J.

Recommendation of Educational Resources in a Blended Learning Environment.

DOI: 10.5220/0011689500003470

In Proceedings of the 15th International Conference on Computer Supported Education (CSEDU 2023) - Volume 1, pages 15-24

ISBN: 978-989-758-641-5; ISSN: 2184-5026

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

study carried out by Ali (2019): increased academic

performance and social skills of students, increased

flexibility in teaching and learning, as well as

increased student satisfaction. There is also a

decrease in dropout rates and an increase in school

retention.

According to Graham (2006), there are three

frequently mentioned definitions concerning blended

learning: à combination of teaching modalities, a

variety of teaching methods, and à combination of

face-to-face instruction with computer-mediated

instruction. It then shows that the first two definitions

are so comprehensive that they would encompass all

education systems. Graham then defines blended

learning systems as combining face-to-face and

computer-mediated instruction. A broader definition

where the expression blended learning is strongly

linked to blended education, in the sense that there is

no single way of learning and in which learning is a

continuous process, occurring in different ways, in

different spaces. This approach expands opportunities

for unified work between these areas.

Recommender systems are systems that generate

suggestions of interesting items for users to use, with

the greatest benefit of reducing the overload of

information on the user, making the interaction of the

user with the system a more attractive activity. and

personalized, considering that the suggestions guide

users towards the existing options (Ricci et al., 2015).

Consequently, other benefits can be pointed out in the

use of recommendation systems, such as decision-

making support, customer loyalty, and increased

revenue, among others (Jannach and Adomavicius,

2016). The items to be recommended can be the most

varied, such as movies, music, books, videos, etc. The

development of recommendation systems has been

the object of companies in order to please and retain

their customers, such as Netflix when recommending

a series, YouTube when recommending a new video,

or even recommending friendships through

Facebook. The work of Ko (2022) points out that

about 21% of the papers published in the last ten years

on recommender systems are intended for use in

social networking systems. Education appears as a

growing purpose, with about 12% of publications.

Thus, the objective of this work is to present a

development of a semantic model for context-aware

recommending educational resources applied to

blended learning environments..

This work is organized as follows: in addition to

this introduction, section 2 presents some related

works; section 3 presents the approach developed;

(http://nutch.apache.org).

section 4 presents the ontology developed; in section

5 a case study is presented through a scenario of use

to evaluate the proposal developed; finally, section 6

presents the conclusions and future work.

2 RELATED WORKS

In the work by Hoic-Bozic et al. (2015), the

implementation of a blended learning environment

was presented using the ELARS system (E-Learning

Activities Recommender System) to recommend

educational resources. The ELARS system

recommends four types of e-activities: optional

activities, potential collaborators (other students),

web 2.0 tools, and mentoring. Wonoseto and

Rosmansyah (2017) developed a knowledge-based

recommendation system applied to a school in

Indonesia. Learning style knowledge and

collaborative learning theory were considered,

performing the modeling in two stages: identification

of learning styles and recommendation based on

learning style and collaboration.

To identify learning styles, based on Flaming's

theory, the authors applied a VAK (visual, auditory,

or kinaesthetic) questionnaire (Visual, Auditory, and

Kinaesthetic) in order to model students based on

VAK learning styles. Then, in the second step, the

learning objects were grouped into characteristics to

be recommended to students according to their

learning styles. Each student also received an

indication of a collaborator with the same learning

style, but with a difference in school performance.

A blended learning model supported by

recommendation systems was proposed by Saied and

Nasr (2018), who performed a recommendation based

on the correspondence between the student and the

recommended resource. The authors performed two

modelings: group modeling and content modeling.

For the group modeling, a two-level collaborative

filtering approach was applied in order to group the

students according to the similarities and differences

between their preferences, while in the content

modeling, indexing, and text mining techniques were

used, using Nutch, an open-source search engine

. As

a recommendation strategy, content-based filtering

was adopted, by mining student models, mining

association rules, tracking and indexing learning

resources, and extracting user preferences.

In the work of Bouihi and Bahaj (2019),

architecture was proposed for recommending

learning objects for distance learning. Based on the

CSEDU 2023 - 15th International Conference on Computer Supported Education

classic 3-layer architecture (presentation, business,

and persistence), it was proposed to insert a semantic

layer between the business and persistence layers.

The proposed layer is composed of two semantic

subsystems: an ontology-based system and an SWRL

rule-based system. The ontology-based subsystem is

used to model the content and context of the learning

object, in order to make it reusable and shareable. The

rule-based subsystem, in turn, is used for

recommendations, made based on the relevance of the

learning object. These rules have been organized into

four categories: Learning History Rules, Learning

Performance Rules, Social Network Learning Rules,

and Learning Path Rules. The authors defined values

5, 10, and 20 as relevance weights for recommending

learning objects (low, medium, and high). These

values are updated with each run of the SWRL rules.

Mendes et al (2017) presented SADE (Student

Performance Monitoring System), a system that

makes recommendations for learning objects based

on student performance profiles. The modeling of the

student's profile is carried out in two stages: first, their

personal data are collected in a registration form;

then, during use, the system collects information from

the student's performance in the evaluations carried

out by the teacher, classifying it in 5 levels (A, B, C,

D, and E). For recommendation, the content-based

technique was used, making use of information from

the student's database as a function of their

development in the assessments.

Tarus et al (2018) proposed a hybrid

recommendation approach based on context

awareness and sequential pattern mining to

recommend learning resources to students in ODL

environments. The authors used context awareness to

insert contextual information about the student, such

as knowledge level and learning objectives. With the

use of contextual information, Collaborative Filtering

was used in the recommendation strategy, through

Pearson's correlation coefficient and the kNN (k

Nearest Neighbour) algorithm. To filter the

recommendations according to the access patterns,

sequential mining of patterns was used, through the

GSP (Generalised Sequential Patterns) algorithm in

order to mine the web logs.

In Jeevamol and Renumol (2021), an ontology-

based e-learning content recommendation system

was proposed to solve the cold start problem. The

proposal is composed of 3 main components:

interfaces, ontologies, and unit of measure of

similarity of student and learning object. In the

ontology component, 3 ontologies were built: student,

to model personal information and learning style

according to FSLSM (Felder - Silverman Learning

Style Model); student log, to model the student

learning path and; material, to model the learning

objects, from the LOM standard. In the similarity

measurement unit, ontology is used to generate

recommendations under cold start conditions,

combined with collaborative and content-based

filtering techniques.

In Harrathi et al. (2017), a proposal for a learning

activity recommendation system was presented in the

context of MOOC (Massive Open Online Course),

with the objective of helping the student to follow the

learning process and reduce evasion. A knowledge-

based recommendation was used as a technique to

carry out the recommendation of distance activities,

based on ontology. The system uses ontology to

model and represents domain knowledge, student,

and learning activities.

The architecture is composed of 3 layers: user

interface, system operation, and database. The User

Interface layer is responsible for sending the data to

feed the student model module and presenting the

recommendations coming from the system operation

layer. The system operation layer contains 4 modules:

student model module, domain model module,

recommendation generation module, and

recommendation display module. The database layer

of the system abstracts 3 databases used by the

system: the student model database, the rules

database, and the learning activities database. The

domain model was based on an ontology related to

the IMS-LD specification to specify the structure of a

MOOC course. The student profile is composed of 4

subclasses: knowledge level; student characteristics;

learning style, and; preferences. The ontology of

learning activities has 6 subclasses: difficulty, type,

start time, duration, and priority.

Labib et al (2017), focused on distance learning,

presented an ontology to model the student profile

based on the creation of interconnections between the

different dimensions of the learning style model,

learning style, and the relevant characteristics of the

student. The authors selected the most used learning

style theories in recent years and surveyed how some

student characteristics relate to each learning style

that has some compatibility. The theories used were:

Gregorc's Mind Styles model, Honey and Munford's

model, Riding Styles, Myer Briggs Styles, Felder

Silverman Styles, and Kolb Styles. The dimensions

explored in the ontology were: decision,

understanding, access mode, lifestyle, organization,

and process.

The authors used the On-to-knowledge

methodology, presented in Sure et al (2009), which

has 5 phases. In phase 1 - Feasibility Study,

Recommendation of Educational Resources in a Blended Learning Environment

information was collected on scientific databases

related to learning styles models. In phase 2 -

Departure, the process of extracting the

characteristics of the student was carried out. In phase

3 - Refinement, 3 sub-phases were performed:

definition of a base taxonomy to formulate the

application-oriented ontology; elicitation, where

many knowledge entities are defined and;

formalization, where entities are organized into

hierarchies. In phases 4 and 5 - Evaluation and

Maintenance, after using the reasoner, several

example queries were applied to test the consistency

and verify the usefulness of the ontology.

Obeid et al (2018) presented a proposal for an

ontology-based recommendation system for

undergraduate students. The approach uses improved

ontology with machine learning techniques to guide

students in higher education, in order to assess

vocational strengths and weaknesses, student

interests and capabilities, in order to identify students'

interests, requirements, preferences, and ability to

recommend the appropriate course and university.

The authors conducted the survey among French and

Lebanese students through university portals.

The proposed system has four main parts: 1 -

explicit and implicit data collection, through the data

filled in their profile (explicit) and the proposal of a

survey to the students to collect information about

their interests (explicit); 2 - ontology, which supports

the system by modeling domain knowledge, through

the creation of three ontologies, educational

institution, job, and student; 3 - machine learning,

which processes information, creating and grouping

student profile models and sending the results to the

hybrid recommendation engine; 4 - recommendation,

through compatibility between students and interests,

performs recommendations and saves student

recommendation information for later use.

Ouf et al (2017) carried out a detailed study on

works that use ontology in teaching environments,

verifying that most studies are focused on

personalizing learning objects, ignoring other factors.

The authors then propose a framework for distance

learning environments, using ontology and SWRL.

The proposal's architecture is composed of four

layers: 1 - Interface, which presents the results of the

recommendations; 2 - Semantic Reasoning Engine,

responsible for personalization through the use of

reasoners, using SWRL; 3 - Semantic Layer, which

contains the semantic representations of the

components of the learning process and; 4 - Semantic

Metadata, notes the concepts and their relationships.

The approach contains four ontologies, which

form the semantic layer: student model, learning

objects, learning activities, and teaching methods.

The student model has several categories: personal,

knowledge, behavioral, preferences, learning

objectives, and safety. The ontology of learning

objects consists of different modules. A module

includes a set of domains, and each domain is made

up of one or more subjects, which include knowledge

of different subjects associated with the student. The

ontology of learning activities defines activities such

as brainstorming, case study, and group work, among

others. The ontology of teaching methods defines

methods such as reflection, project-based learning,

and workshops, among others.

Agarwal et al (2022) proposed a hybrid

recommender system that utilizes cluster-based

collaborative filtering and rule-based

recommendation using SWRL for recommendations

applied to MOOC platforms in order to recommend

internal course elements along with the learning path

in addition to learning tips. apprenticeship.

To carry out the grouping of students, the use of

the learning style Felder Silverman's Learning Style

Model (FSLSM) was incorporated through the

detection of traced usage parameters. The authors use

contextual knowledge to deal with the cold-start

problem and SWRL to perform rule-based

recommendations with ontologies. Two ontologies

were used: learner and course. The learner ontology

is used to represent the static (such as SHA-256 key,

platform used, browser, etc.) and dynamic (courses

currently enrolled, learning style, grades, return

visits, the fraction of completion, etc.) attributes of

the learner. The course ontology also contains static

and dynamic attributes. Static attributes include the

course name, course duration, and course fee, along

with the set of all course elements regarding video

lectures, reading material, etc. Dynamic attributes

include current enrollment, current completion, etc.

Then, after grouping according to the FSLSM

learning style, the model makes recommendations

through SWRL rules.

Ezaldeen et al (2022) proposed a framework

called Enhanced e-Learning Hybrid Recommender

System (ELHRS) for E-learning recommendation

that integrates adaptive profiling and sentiment

analysis.

The authors developed a semantic model to

deduce the learner's profile automatically and used

the DBPedia (http://www.dbpedia.org/) and WordNet

(https://wordnet.princeton.edu/) ontologies as a

semantic-based approach for term expansion so that

the learner's profile is updated according to their

behavior and different navigation actions. . As part of

the recommendation system, custom constructions of

CSEDU 2023 - 15th International Conference on Computer Supported Education

Convolutional Neural Networks (CNN) and the

hybridization of Natural Language Processing (NLP)

methods were developed by thirteen CNN-based

models of sentiment analysis to predict the ratings of

textual reviews on a resource specific learning.

It is noted that there are many research efforts in

the sense of recommending learning resources that

perform the modeling of the learner profile, such as

the work of Mendes et al (2017). Some works use

semantic web and ontology, such as the works by

Wonoseto and Rosmansyah (2017), Jeevamol and

Renumol (2021), Harrathi et al. (2017), Obeid et. al

(2018), Ouf et al (2017) and Ezaldeen et al (2022).

The works by Hoic-Bozic et al. (2015) and Saied and

Nasr (2018) address hybridism, while the works of

Tarus et al. (2018) and Agarwal et al (2022) use

context awareness, but without addressing hybridism.

Thus, the present work is able to collaborate with

what has already been done by exploring the

sensitivity to the student context applied to the hybrid

education environment, by modeling physical and

digital learning resources, with their respective usage

requirements and the student context in the location

and technology dimensions, which make it possible

to enrich the recommendation process.

3 DEVELOPED APPROACH

The problem of information overload, already known

in educational environments, is potentiated in blended

learning environments, because, in addition to the

numerous options that the student has to deal with to

choose an educational resource that helps him/her in

the teaching process, there are specific limitations due

to the context in which the student lives and is at the

moment. For example, the student may have a

residence with high-speed connection support and a

high-processing computer, or a residence with no

connection, and only have a smartphone with a

mobile data connection. This work proposes an

ontological model for recommending learning

resources applied to blended learning.

In everyday life, an example that can be used is a

student who is at home and wants to use a learning

resource to help him/her perform a task for the next

week. It has numerous search options, and the search

for resources does not take into account your

conditions of use. This generates what is called

cognitive overload and reduces the efficiency of your

study.

To deal with this problem, a model was developed

that includes aspects of the user profile (general and

academic), in addition to considering three

dimensions of context: technology, location, and data

from learning resources. Thus, it is expected to help

the student in the process of choosing the available

resources, reducing the time of searching for them,

and helping the learning process. The model will be

further detailed in the next subsections, with the

presentation of the modeling of each of the contextual

aspects. There are several ways to model a domain;

that is, there is not only one alternative considered

correct. The methodology used in the construction of

the ontological model proposed in this work was the

“Ontology development 101”, proposed by Noy and

McGuinnes. The methodology proposed by the

authors presents a simple and interactive way, as step-

by-step, for the construction of an ontology, which

justifies its choice in this work. The process is divided

into seven steps to be followed during model

development.

4 DEVELOPED ONTOLOGY

Protégé (Stanford, 2022) stands out as the most used

ontology editor by the international community. In

this work, Protégé version 5.5.0 was used to build the

ontology. Figure 1 presents the overview of the model

developed, with the information aspects used to

achieve the recommendations of learning resources to

the student. The student is the center of the model,

which has two profiles: general and academic. It

relates to the aspects of location, technology, and

subject of interest, in order to have the recommended

learning recommendation. Technology refers to the

technological aspects that concern the student, such

as the type of connection, the screen size of the

equipment, the presence or absence of GPS, and

processing capacity. Location describes the student's

physical location perspective.

The location has spatial points that describe it

(latitude and longitude) and can be automatically

captured by the student's device or manually entered

depending on the context. In the model, specific

subclasses were described for the purpose of the

recommendation, which are: home, campus, and

academic environment.

The learning resource is the class that defines the

objects that will be recommended to the student based

on their profile, and their relationship with aspects of

technology and location. Two subclasses were

modeled: physical learning resources and digital

learning resources. This definition is important

because, in a blended learning environment, this

factor is decisive for the recommendation or not of a

Recommendation of Educational Resources in a Blended Learning Environment

certain resource at a given moment of the student's

study.

Figure 1: General Model Classes Developed.

The general model defines the central concepts of

the ontology, with which all the other classes are

related, which is shown in Figure 1. Some object

properties were omitted to allow better visualization

of the classes.

The Learner class relates to the Profile class

through the hasProfile object property.

AcademicProfile and GeneralProfile are subclasses

of the Profile class, which model, respectively, the

Academic Profile and the General Profile of the

student. It is worth noting that the student relates to a

certain subject, Topic class, through the h-sInterestIn

object property in his GeneralProfile. The

LearningResource class relates to the Topic class

through the coversTopic object property and

has two subclasses DigitalLearningResource

and PhysicalLearningResource. The

PhysicalLearningResource class is related to the

Place class through the hasPlace object property, as

it has a storage location, for example, a Book that is

stored in a library. A Topic can have a prerequisite

that is another Topic, modeled through the

hasPreRequisite object property.

A more detailed relationship between the learner

profile, technology, location, and learning resources

is shown in Figure 2. The overall learner profile

relates to technology through a hasTechnology object

property. On the other hand, the learner's class relates

to technology through the usesTechnology property,

which defines the technology the student uses when

the recommendation is made. This was modeled in

this way since a student who is at home, even using a

smartphone with a small screen, can receive a

recommendation for a resource that requires greater

computing power if he has a computer in his home

that meets the minimum requirements of the resource.

It is also possible to notice the relationship that the

learner has with a location through the hasLocation

object property, which is essential for the

recommendation of physical resources. The

PhysicalLearningResource class is related to the

Location class through the hasStorageLocation

object property, as it has a storage location, for

example, a Book stored in a library. The Location

class also has an object property for defining one

location to be close to another, isNearOf, which can

also be used to make a recommendation.

Figure 2: Relationship between models of the learner,

location, technology, and learning resources.

5 CASE STUDY

A case study is presented in order to evaluate the

approach developed in this work. A case study is

characterized as an empirical investigation that

investigates a contemporary phenomenon in depth in

its real-world context, especially when the boundaries

between the phenomenon and the context may not be

so clearly evident. In this sense, the case study is

developed from the possibilities of use of the

proposal, where recommendations made from the

student's profile and the context in which he is, within

a blended learning environment are reported. First,

the possibility of using the developed model is

presented. Then, a presentation is made on the

inference mechanisms used in the proposal. Finally, a

usage scenario is presented with the application of

SWRL rules, in the situation of use by the student

from his/her house.

In a blended education environment, there is a

demand for recommending educational activities to a

student based on their characteristics, knowledge, and

context. As for their characteristics, the student can

have a general profile and an academic profile. In the

academic profile, the educational characteristics of

the student are considered, such as previous

knowledge of a certain subject and the student's

learning objectives. In the general profile, the

characteristics that the student has, regardless of their

school situation, are taken into account, such as

CSEDU 2023 - 15th International Conference on Computer Supported Education

housing, subjects of interest, and work. The system

also stores the student's knowledge through the

subjects with which he has already interacted and the

classes he has attended.

To perform the filtering that makes the

recommendation of learning resources effective,

SWRL rules are used. SWRL was proposed in

Horrocks et al. (2004) as an OWL syntax extension

language by combining the OWL DL and OWL Lite

sublanguages with RuleML. Thus, SWRL extends

OWL's semantic representation capacity through

first-order logic rules, being recommended by the

W3C for this purpose. The rules are constituted by

antecedent => consequent, where, from the

satisfaction of the antecedent's criteria, the

consequent is generated. To execute the SWRL rules,

an inference engine, or reasoner, is used. There are

several reasoners available for use and compatible

with the Protégé software, used in this work, among

which the following stand out: FaCT ++, HermiT,

ELK, Pellet, and Racer. In the construction of the

ontology developed in this work, we chose to use the

Pellet, because, in addition to being Open Source, it

offers a set of functionalities such as rule support,

total incremental classification, and greater

expressiveness of descriptive logic combining two

native profiles.

“

The scenario considers Johan, a student

of the Computer Engineering course, which is

taught in the hybrid learning mode. The course

program adopted the enriched virtual model;

that is, the course mostly takes place online,

with some regular face-to-face meetings. In

the Introduction to Artificial Intelligence

discipline, after the presentation of the

content, the teacher requests that a

practical activity to implement the k-means

algorithm be carried out, which will have its

conclusion and debate in the next face-to-

face meeting. After the teacher enters the

class in the system with the subject and the

task to be performed, "k-means" is added as a

subject of interest to Johan.

In the next access to the system, it

appears that Johan is accessing through a

mobile device and that he is at his residence.

Johan does not have a computer with high

processing capacity, and his connection is

not high-speed. From there, the system infers

the recommendation rules in order to return

some suggestions of learning resources in

order to help Johan in carrying out the task

”.

In Figure 3, it is possible to see how the learning

resources related to the student profile and the

characteristics of the technology in use. Within the

model, several aspects can be taken into account to

make the recommendation. In this case study, it is

considered that if the student is at home, only digital

learning resources will be presented.

Figure 3: Model Instantiated with a student at home.

Since digital resources have minimum

requirements defined for their good use, satisfaction

with the technology available to the student will also

be considered, in order to recommend only those that

can be used effectively. For this, having the device

information available to the student, the system

checks what its processing standard is and what is the

minimum processing requirement for each digital

learning resource. Then, it does the comparison

through the built-in function swrl:lessThanOrEqual(),

to check if the processing requirement is less than or

equal to the device's processing standard. The same

process is repeated to check the screen size and

connection patterns. Finally, the relationship between

the learning resource and the student's profile is

demonstrated through the correspondence between

the subject of interest to the student and the scope of

the resource to be recommended. The SWRL rule

developed to effect the above recommendation is

described below:

Learner(?l) ^

House(?h) ^

Profile(?p) ^

Topic(?topic) ^

DigitalLearningResource(?resource) ^

hasProfile(?l, ?p) ^

hasLocation(?l, ?h) ^

usesTechnology(?l, ?technology) ^

hasConnectionPattern(?technology,

?connection) ^

hasConnectionPatternRequirement(?res

ource, ?connectionRequirement) ^

swrlb:lessThanOrEqual(?connectionReq

uirement, ?connection) ^

hasProcessingPattern(?technology,

?processing) ^

Recommendation of Educational Resources in a Blended Learning Environment

hasProcessingPatternRequirement(?res

ource, ?processingRequirement) ^

swrlb:lessThanOrEqual(?processingReq

uirement, ?processing) ^

hasScreenSizePattern(?technology,

?screenSize) ^

hasScreenSizePatternRequirement(?res

ource, ?screenSizeRequirement) ^

swrlb:lessThanOrEqual(?screenSizeReq

uirement, ?screenSize) ^

hasInterestIn(?p, ?topic) ^

coversTopic(?resource, ?topic) ->

hasRecommendedLR(?l, ?resource)

Figure 4: Result of the Rules Inferred by Pellet in the case

study.

Figure 4 shows an image of the Protégé screen

where the result of the recommendation is displayed.

The object properties in yellow are those obtained by

reasoning about the presented rule, used by the Pellet

inference engine, from the model instance.

It is noted that from the characteristics presented

(location, technology used, and learner interest), the

model was able to infer that the learner Johan will

have the digital article Means Clustering and the slide

presentation KMeans Presentation recommended in

order to assist in the execution of your task of

implementing the k-means algorithm, within the

discipline of Introduction to Artificial Intelligence.

An evaluation of the model developed through the

use of a case study was carried out, with the

presentation and use scenario, where situations were

simulated in which a student in the hybrid learning

modality is using a system developed from this model

to receive recommendations of learning resources in

different contexts. In the scenario, a student is at

home and receives recommendations from the

technology available for study. In this case, the

student's interest is considered, based on the modeled

profile.

The advantage of using the knowledge-based

recommendation technique is noted, since, from the

semantic descriptions, it is possible to obtain better

results in a situation of new users, as they can receive

recommendations based on their location and the

technology in use, which helps to solve the cold start

problem. Another advantage is that by using

application semantics at a higher conceptual level,

domain knowledge becomes understandable by

computational and human agents.

6 CONCLUSIONS

In this paper, an evaluation of the model developed

through the use of a case study was carried out, with

the presentation of a scenario of use, where a situation

were simulated in which a student in the blended

learning modality is using a system developed from

this model to receive recommendations of learning

resources in different contexts. In that scenario, a

student is at home and receives recommendations

considering the technology available for study. In

that case, the student's interest is considered, based on

his modeled profile. The advantage of using the

knowledge-based recommendation technique is

noted, since, from the semantic descriptions, it is

possible to obtain better results in a situation of new

users, as they can receive recommendations based on

their location and the technology in use, which helps

to solve the cold start problem. Another advantage is

that by using application semantics at a higher

conceptual level, domain knowledge becomes

understandable by computational and human agents,

as suggested by Berners-Lee et al (2001).

This work presents the development of an

ontology for recommending learning resources in

context-sensitive blended learning environments. The

main context dimension taken into account was

location since, in blended learning, the student's

residence becomes an extension of the school

environment. Thus, physical and digital learning

resources were modeled, to be recommended

differently in a situation of use in an academic

environment or in a residence. Another context

dimension considered was the technology that the

student has, and that is currently in use so that

learning resources can be filtered according to the

technological capabilities available to the student.

In future works, we intend to generate a module

and integrate it into a Virtual Learning Environment.

Subsequently, it is intended to carry out the validation

through tests with students in a real environment of

blended learning. Another future work will be to

model different user and context characteristics, such

as cultural issues., increasing the personalization of

the recommendation. Finally, we intend to insert

other context dimensions, such as activity and time.

CSEDU 2023 - 15th International Conference on Computer Supported Education

ACKNOWLEDGEMENTS

This research is supported by CNPq/MCTI/FNDCT

Nº 18/2021 grant n. 405973/2021-7 and CAPES -

Financing Code 001. The research by José Palazzo M.

de Oliveira is partially supported by CNPq grants

306695/2022-7 PQ-SR. The research by Isabela

Gasparini is partially supported by CNPq grant

308395/2020-4 and FAPESC Edital nº027/2020 TO

n°2021TR795.

REFERENCES

Agarwal, Anant. The future of learning is blended.

http://hdl.handle.net/1853/64299. In: Moving

Horizontally: The New Dimensions of At-Scale

Learning at the Time of COVID-19, edited by Yakut

Gazi and Nelson Baker.

Agarwal, Abhinav; Mishra, Divyansh Shankar; Kolekar,

Sucheta V. Knowledge-based recommendation system

using semantic web rules based on Learning styles for

MOOCs. Cogent Engineering, v. 9, n. 1, p. 2022568,

2022.

Ali, Ramiz. Is blending the solution?: a systematic literature

review on the key drivers of blended learning in higher

education. 2019.

Anthony, Bokolo et al. Blended learning adoption and

implementation in higher education: a theoretical and

systematic review. Technology, Knowledge and

Learning, p. 1-48, 2020.

Berners-Lee, Tim; Hendler, James; Lassila, Ora. The

semantic web. Scientific American, v. 284, n. 5, p. 34-

43, 2001.

Bouihi, Bouchra; Bahaj, Mohamed. Ontology and Rule-

Based Recommender System for E-learning

Applications. International Journal of Emerging

Technologies in Learning, v. 14, n. 15, 2019.

Bruggeman, Bram et al. Experts speaking: Crucial teacher

attributes for implementing blended learning in higher

education. The Internet and Higher Education, v. 48, p.

100772, 2021.

Ezaldeen, Hadi et al. A hybrid E-learning recommendation

integrating adaptive profiling and sentiment analysis.

Journal of Web Semantics, v. 72, p. 100700, 2022.

Graham, Charles R. Blended learning systems. The

handbook of blended learning: Global perspectives,

local designs, v. 1, p. 3-21, 2006.

Harrathi, Marwa; Touzani, Narjess; Braham, Rafik. A

hybrid knowledge-based approach for recommending

massive learning activities. In: 2017 IEEE/ACS 14th

International Conference on Computer Systems and

Applications (AICCSA). IEEE, 2017. p. 49-54.

Hoic-Bozic, Natasa; Dlab, Martina Holenko; MORNAR,

Vedran. Recommender system and web 2.0 tools to

enhance a blended learning model. IEEE Transactions

on education, v. 59, n. 1, p. 39-44, 2015.

Horrocks, Ian et al. SWRL: A semantic web rule language

combining OWL and RuleML. W3C Member

submission, v. 21, n. 79, p. 1-31, 2004.

Jannach, Dietmar; Adomavicius, Gediminas.

Recommendations with a purpose. In: Proceedings of

the 10th ACM conference on recommender systems.

2016. p. 7-10.

Jeevamol, Joy; Renumol, V. G. An ontology-based hybrid

e-learning content recommender system for alleviating

the cold-start problem. Education and Information

Technologies, v. 26, n. 4, p. 4993-5022, 2021.

Ko, H.; Lee, S.; Park, Y.; Choi, A. (2022). A survey of

recommendation systems: recommendation models,

techniques, and application fields. Electronics, 11(1),

141.

Labib, A. Ezzat; Canós, José H.; Penadés, M. Carmen. On

the way to learning style models integration: a Learner's

Characteristics Ontology. Computers in Human

Behavior, v. 73, p. 433-445, 2017.

Mendes, Tiago de Avila et al. A Recommendation Method

of Didactic Content to Accompany Student

Performance. INTED2017 Proceedings, 2017.

Obeid, Charbel et al. Ontology-based recommender system

in higher education. In: Companion Proceedings of The

Web Conference 2018. 2018. p. 1031-1034.

Oliver, Martin; Trigwell, Keith. Can ‘blended learning’ be

redeemed?. E-learning and Digital Media, v. 2, n. 1, p.

17-26, 2005.

Ouf, Shimaa et al. A proposed paradigm for a smart

learning environment based on semantic web.

Computers in Human Behavior, v. 72, p. 796-818,

2017.

Rafiola, Ryan et al. The Effect of Learning Motivation,

Self-Efficacy, and Blended Learning on Students’

Achievement in The Industrial Revolution 4.0.

International Journal of Emerging Technologies in

Learning (iJET), v. 15, n. 8, p. 71-82, 2020.

Ricci, F., Rokach, L., and Shapira, B. (2015).

Recommender systems: introduction and challenges. In

Recommender systems handbook, pages 1–34.

Springer.

Rooney, J. E. Blending learning opportunities to enhance

educational programming and meetings [Tekst].

Association Management [Tekst], n. 55, p. 5, 2003.

Saied, Mohamed; Nasr, Mona. Blended learning model

supported by recommender system and up-to-date

technologies. International Journal of Advanced

Networking and Applications, v. 10, n. 2, p. 3829-3832,

2018.

Stanford. Protégé Ontology Editor. http://protege.

stanford.edu/. Accessed: 2022-11-30

Sure, York; Staab, Steffen; Studer, Rudi. Ontology

engineering methodology. In: Handbook on ontologies.

Springer, Berlin, Heidelberg, 2009. p. 135-152.

Tarus, John K.; Niu, Zhendong; Kalui, Dorothy. A hybrid

recommender system for e-learning based on context

awareness and sequential pattern mining. Soft

Computing, v. 22, n. 8, p. 2449-2461, 2018.

Vo, Minh Hien; Zhu, Chang; Diep, Anh Nguyet. Students’

performance in blended learning: disciplinary

Recommendation of Educational Resources in a Blended Learning Environment

difference and instructional design factors. Journal of

Computers in Education, v. 7, n. 4, p. 487-510, 2020.

Wonoseto, M. G., & Rosmansyah, Y. (2017, October).

Knowledge based recommender system and web 2.0 to

enhance learning model in junior high school. In 2017

International Conference on Information Technology

Systems and Innovation (ICITSI) (pp. 168-171). IEEE.

CSEDU 2023 - 15th International Conference on Computer Supported Education