Teacher Educational Resources Recommendation in the COVID-19

Context

Nader N. Nashed

1,2 a

, Christine Lahoud

2 b

and Marie-H

ene Abel

1 c

HEUDIASYC, Universit

e de Technologie de Compi

egne, Compi

egne, France

Universit

e Franc¸aise d’

Egypte, Cairo, Egypt

Keywords:

Education, Recommender System, Semantic Web, Ontology, Sentimental Impact, COVID-19.

Abstract:

The educational process is greatly impacted by the coronavirus pandemic (COVID-19) and this impact out-

reaches teachers, learners, and all participants. This pandemic is responsible for multiple modiﬁcations to the

traditional teaching and learning techniques and technologies which leads to social and cultural changes. Due

to the accompanied social changes, teachers are experiencing different degrees of stress, burnout, and work

difﬁculties during learning new technologies and preparing to their courses’ contents. Teacher’s sentimental

state can be represented by one’s mood and can present an accurate measurement to a teacher experiencing the

mentioned difﬁculties. Accordingly, this study proposes a possible solution for one of the main participants in

the educational process, the teacher, by recommending him educational resources to cope with stress and pre-

vent burnout. The proposed recommender system analyses the teacher’s mood and accordingly recommends

educational resources that can enhance the teacher’s sentimental state. Through the investigation of this pro-

posal, we found that the enforcement of the sentimental state impacts the resulting recommendations that are

presented to teachers.

1 INTRODUCTION

During the current ever-changing circumstances, the

educational process, along with everything, is experi-

encing signiﬁcant adjustments and adaptations in fa-

vor of learners to cope with this coronavirus pandemic

(COVID-19) span. The most signiﬁcant adjustment

to the educational system is the rapid shifting towards

online teaching and learning environments (Peimani

and Kamalipour, 2021). The online teaching/learning

environments form a major challenge for the educa-

tional process, as a result, participants lose control

over the participation hours which greatly affects their

psychological and sentimental states (Akour et al.,

2020). Therefore, teachers, like many professions,

are forced to have an unstable work-life balance due

to shifting towards teleworking (Thulin et al., 2019).

Stress and burnout are closely associated with the in-

tensive usage of new technologies by teachers as a

cause of lack of previous adaptation to these technolo-

gies (Riedl et al., 2012).

https://orcid.org/0000-0001-5849-1322

https://orcid.org/0000-0002-4520-634X

https://orcid.org/0000-0003-1812-6763

Prior to the coronavirus pandemic, teachers expe-

rience multiple struggles including stress and work

overload (Nashed et al., 2022). During the pandemic,

these struggles have been highlighted and intensiﬁed

by the lack of training and the inequality of access

to modern technology in the different regions (Jain

et al., 2021; Scherer et al., 2021). (Friedman, 2000)

highlighted three components of burnout: sentimen-

tal exhaustion, detachment from the job, and a lack

of personal accomplishment. With the proper train-

ing, teachers can overcome these causes and prevent

the psychological risk accompanied with health cri-

sis (Jenaro et al., 2007). This training develops the

professional capabilities of teachers and accordingly,

increases their self-esteem and eliminates the feel of

guilt (Jenaro et al., 2007). Training is considered as

a type of educational resources along with academic

improvement activities, skills in a speciﬁc subject, in-

novative teaching methodologies, stress management

at work, time management, or even one or more peers

for work collaboration and experience sharing. More-

over, the multiple contexts in which any teacher co-

exists must be considered to provide the teacher with

proper training (Cooper and Olson, 2020).

As a result, this pandemic enforces the considera-

Nashed, N., Lahoud, C. and Abel, M.

Teacher Educational Resources Recommendation in the COVID-19 Context.

DOI: 10.5220/0011147000003182

In Proceedings of the 14th International Conference on Computer Supported Education (CSEDU 2022) - Volume 1, pages 587-598

ISBN: 978-989-758-562-3; ISSN: 2184-5026

587

tion of one’s sentimental context as one of the indica-

tions to the stress and burnout (Herman et al., 2018;

Skaalvik and Skaalvik, 2020). During previous re-

search, the teacher’s multiple contexts were identi-

ﬁed, and their impact was discussed (Flores and Day,

2006). These contexts comprise the living environ-

ment context as well as the working environment con-

text to shape the major contextual factors that affect

the teacher’s performance. Mood, emotional commit-

ment, and ﬂow shape the sentimental context of the

teacher (Bishay, 1996).

In this paper, we explain the usage of merging of

the sentimental context, raised by COVID-19, into the

already existing multiple contexts of teachers to pro-

vide personalized educational resources recommen-

dations. Section 2 provides a background review of

related work to this one. Then, Section 3 contains an

illustration of the proposed methodology. Moreover,

Section 4 discusses the usage of this proposition with

the aid of examples. Finally, Section 5 concludes this

paper and introduces future perspectives.

2 RELATED WORK

The physical and mental health of teachers are

strongly related to their working efﬁcacy (McIntyre

et al., 2017; Lachowska et al., 2018; Oberle et al.,

2020). The resultant mood swings, during the cur-

rent times, are responsible for teacher efﬁcacy and

performance (Frenzel, 2014). In addition, it is di-

rectly related to the teacher’s burnout which directly

affects the learners (M

erida-L

opez and Extremera,

2017; Heutte et al., 2016). Educational/training re-

sources can act as a mitigator to COVID-19 effects

on teachers such as burnout and stress (Lizana et al.,

2021). These educational resources are categorized

into internal resources, such as classroom manage-

ment and instructional resources, and external re-

sources, such as supporting educational resources.

This approach is said to be efﬁcient if teachers are

provided with personalized strategies and resources to

acquire new skills and facilitate the classroom man-

agement (Lizana et al., 2021).

During the current COVID-19 era, digital tech-

nologies are essential support for teaching and learn-

ing processes in various ﬁelds and different educa-

tional levels (Perifanou et al., 2021; Mukhtar et al.,

2020). Educational resources recommender sys-

tems (ERRS) aid this purpose by providing personal-

ized educational resources recommendations for each

teacher. Traditionally, ERRS targets learners by rec-

ommending learning objects or materials in addition

to performance evaluation (Zhang et al., 2021). The

importance of the context-aware recommenders arises

recently to consider the social and environmental con-

ditions (Nilashi et al., 2020). (De Meo et al., 2017)

introduced a social networking context-aware recom-

mender system approach by combining trust relation-

ships and skills evaluation to reform the online class-

rooms. On the other hand, (Kla

snja-Mili

cevi

c et al.,

2018) combines social tagging and sequential patterns

to provide context-aware recommendations in an e-

learning environment. (Moore et al., 2019) monitors

student’s progress in order to provide personalized

performance evaluation without using face-to-face tu-

torial sessions and then recommends learning materi-

als. As for the teacher centered ERRS, (Cobos et al.,

2013) introduced a recommender approach to teach-

ers by detecting the pedagogical patterns from the

online recorded information based on the class con-

text. The systematic reviews conducted by (Zhang

et al., 2021) and (Imran et al., 2021) prove that re-

search efforts are directed towards learners’ context

with nearly neglection to teacher context which was

the motivation of this research.

3 PROPOSED APPROACH

The task of providing personalized recommendations

is a complicate process, moreover, the considera-

tion of teacher’s/user’s context increases the difﬁculty

level of this task. At the beginning, teacher’s multi-

ple contexts are summarized into living environment,

working environment and sentimental state within the

organizational management system, MEMORAe, and

are represented by teacher-context ontology (TCO),

mood detection ontology (MDO), as well as MEM-

ORAe ontology (MCC) (Nashed et al., 2021). Each

context is deﬁned by a teacher-dependent set of fac-

tors and dimensions which needs to be precisely se-

lected in order to avoid the uniﬁcation of these fac-

tors. However, the selection of these factors cannot

grantee accurate representation of teacher’s contexts,

instead, the factors’ weighting is another important

step towards the precise teacher’s contexts represen-

tation. Therefore, the proposed approach, as shown

in Fig. 1, introduces a ﬁrst step of mood detection us-

ing MoodFlow@doubleYou (Andres et al., 2021) and

user activity tracking using the MEMORAe SoIS for

a teacher. Afterwards, the contextual factors are ex-

tracted and weighted with a mood-enforcing approach

with respect to the teacher. These contextual factors

are then used to ﬁnd matching teachers with similar

context to the current teacher which are stored into a

EKM 2022 - 5th Special Session on Educational Knowledge Management

588

Contextual

Factors

Extraction

Weights

Optimization

Teacher

Activity

Tracking

Teacher

Mood

Detection

Context

Matching

Select/Sort

Resources

Figure 1: Overview of the approach architecture.

sorted list according to a set of SWRL rules

. At the

end, a list of educational resources recommendations

is retrieved for this teacher using the list of matching

teachers. Accordingly, this paper proposes a method-

ology to select and weight the contextual factors to

facilitate the educational resources recommendation

for teacher.

3.1 Ontology Representation

The ontology representation of the aforementioned

contexts, as shown in Fig. 2, facilitates the intercon-

nectivity between them, in addition to the data repre-

sentation advantage. TCO represents a teacher as a

person that is related to two types of environments:

living and working environments. Object properties

are used to associate descriptive objects to each en-

vironment such as location, area type, population,

etc. The educational institution, in which the teacher

works, is related to the work environment to facilitate

the working conditions representation.

https://www.w3.org/Submission/SWRL/

While MDO ontology represents the mood of a

certain teacher using the three-level representation of

mood: negative, neutral, and positive moods. The

teacher’s mood is captured while interacting with

educational resources using MoodFlow@doubleYou

technology. The sentimental context is represented

through the associated mood concept with each

tracked activity that is recorded for each user account

concept of MEMORAe’s person.

MCC ontology is an organization-management

ontology that describes the data model of MEMO-

RAe SoIS (Abel et al., 2007). This SoIS is used

to manage sub-ontologies and in our case, teachers

are introduced to an organizational representation of

one of their educational courses in the form of no-

tions/concepts. An example, is shown in Fig. 3,

illustrates the ontology representation of ”PRE103”

course which is an introductory course to program-

ming languages and algorithms. Sections and subsec-

tions of ”PRE103” content are represented by nodes

in a semantic mapping manner where each node acts

as a sharing space regarding a certain topic. This

course is divided into sub-node collections of tuto-

rials (TD) and practical exercises (TP) which allows

the teacher to attach educational resources related to

each of the TD and TP nodes. Teacher’s interactions

with node-attached resources, are tracked by the sys-

tem and are associated with one of the mood levels of

MDO.

This mechanism empowers the data representa-

tion and retrieval of the teacher’s context and assists

the following step of contextual factors selection and

weighting.

3.2 Contextual Factors Extraction and

Teacher Matching

A multi-step approach, as shown in Fig. 1 and Al-

gorithm 1, is proposed to extract contextual factors,

optimize their weights, ﬁnd matching teachers’ con-

texts, and recommend educational resources accord-

ing to the matching teachers list. Table 1 includes the

mentioned abbreviations in this sub-section to help

the reader to follow the multiple algorithms and the

sub-algorithms. At the beginning, Algorithm 1 se-

lects a set of contextual factors for a teacher’s con-

text C

, as shown in Algorithm 3, and computes the

ﬁnal weights for each of these factors, as shown in

Algorithm 2. Then, a new teacher’s context C

is de-

ﬁned with the utilization of the selected factors. Ac-

cordingly, the algorithm retrieves a list of matching

teachers according to the newly deﬁned context, as

shown in Algorithm 4. Afterwards, the obtained list is

Teacher Educational Resources Recommendation in the COVID-19 Context

589

mdo:resultsIn

Mood

MDO

owl:equivalentClass

rdf:is-a

Person

TCO

mcc:hasUserAccount

Person

MCC

UserAccount

MCC

mcc:hasActor

Interaction-Activity

MCC

Teacher

TCO

Living

Environment

TCO

Environment

TCO

Work

Environment

TCO

Educational

Institution

TCO

rdf:is-a

tco:evolovesIn

tco:worksIn

tco:livesIn

Figure 2: Partial Tbox of ontology representation of

teacher’s multiple contexts.

scored and sorted according to a set of SWRL rules,

as shown in Algorithm 5. At the end, the obtained

matching teachers list is used to recommend educa-

tional resources for the current teacher T .

Algorithm 2 illustrates the process of computing

the initial weights W

init

of a teacher context C

by us-

ing the variance-based extraction of contextual factors

for teacher T according to the collected interactions’

history H

with a set of nodes in which each node is

connected to a set of resources R, as shown in Algo-

rithm 3. The initial weights W

init

are optimized using

an optimization algorithm such as particle swarm

optimization (PSO) (Marini and Walczak, 2015)

with the root mean squared error (RMSE) (Equation

1) as the ﬁtness function FF which evaluates the

ﬁnal weights W

f inal

of each selected contextual factor.

RMSE(rating

predict

resr

,rating

actual

resr

) =

∑

i=1

(rating

predict

resr

− rating

actual

resr

)

(1)

where:

• n: number of predicted ratings

• rating

predict

resr

: predicted rating for a resource resr

• rating

actual

resr

: actual rating for a resource resr

Table 1: Abbreviations List.

Abbrev. Deﬁnition

TCO Teacher-Context Ontology

MDO Mood Detection Ontology

MCC MEMORAe Ontology

T a teacher

selected teacher

resr an educational resource

rating a rating provided by teacher T for a

resource resr

F set of contextual factors correspond-

ing to a teacher T

f a contextual factor

work

working environment context of

teacher T

living

living environment context of teacher

sentimental

sentimental context of teacher T

selected context of teacher T

all

list of selected contexts of all teachers

history of teacher T

w the weight of a contextual factor f

init

initial weights of contextual factors F

for a teacher T

f inal

ﬁnal weights of contextual factors F

for a teacher T

FF ﬁtness function

coe f

mood

mood contextual factor enforcing co-

efﬁcient

thr

similarity threshold

thr

weight threshold

thr

sim

semantic similarity threshold

thr

score

SWRL rules score threshold

number of similar teachers for a cer-

tain teacher T

Sim

total

total similarity score

sim

resr

similarity score for a single resource

resr

sim

resr

var

similarity variance for a single re-

source resr

SW RL

set of SWRL rules

IC Information Content

LCS Least Common Subsumer

Algorithm 3 shows the process of selecting the

contextual factors from the multiple contexts of a

teacher: C

work

, C

living

, and C

sentimental

. The same

steps are performed in accordance with a resource

resr foreach context of teacher T . First, we calcu-

late the similarity score for each teacher who interacts

with the same resource resr and then, we eliminate

teachers with scores less than the similarity thresh-

old thr

. In order to highlight the sentimental con-

text during this pandemic era, a coefﬁcient multiplier

EKM 2022 - 5th Special Session on Educational Knowledge Management

590

Arrays

Memory

String

One-dimentional Array

Basic Types

Constants and Variables Definition

Control Structures

Loop Instructions

Return Instructions

Control Instructions

Branch Instructions

Introduction to Target Language

Basic Inputs and Outputs

General Struture

Lexical Elements

Different Langauage Operators

Expressions

Fundamental Types

Program Notions

Computer Programming

Algorithm General Structure

Algorithm Definition

Algorithm Schematic Representation

Algorithm Properties

Basic Actions and Datatypes

Basic Types Operators

Input/Output Instructions

Assignement Operator

TD01

TD03

TD04

TD06

TD07

TD08 TD09

TD10

TD11

TD12

TD02

TD05

TP01

TP03

TP04

TP06

TP07

TP08

TP09

TP10

TP11

TP12

TP02

TP05

PRE103

Figure 3: Partial screenshot of ”Introduction to programming course (PRE103)” within the MEMORAe system’s user inter-

face.

Algorithm 1: An overview of the proposed algorithm.

Input: C

work

: current working environment context of a teacher T

living

: current living environment context of a teacher T

sentimental

: current sentimental context of a teacher T

: history of current teacher T

Output: sortedList : sorted list of all matching teachers

begin

1: W

f inal

= contextsFactorsExtraction(C

work

, C

living

, C

sentimental

, H

);

2: C

= getTeacherContext(W

f inal

);

3: MatchingTeacherList = teachersContextMatching(C

, W

f inal

);

4: sortedTeacherList = SWRLsortList(MatchingTeacherList);

5: resourcesList = getRecommendations(H

,sortedTeacherList);

6: return sortedList;

end

is used with the mood contextual factor to enforce

the teacher-teacher context similarity which results in

higher probability of mood factor selection within the

ﬁnal set of factors. Afterwards, we compute the mean

value between the average similarity across similar

teachers interacting with the resource resr and the

variance value within this list to calculate the initial

weight w

init

for this resource resr. If w

init

is less than

the weight threshold thr

, the corresponding factor

will be eliminated from the ﬁnal set of selected con-

textual factors which is used to construct the ﬁnal list

of matching teachers.

Through Algorithm 4, this context matching

step is responsible for ﬁnding the list of teachers

with the same set of selected contextual factors

matching the current teacher’s set of contextual

factors C

by calculating the Jiang-Conrath semantic

similarity between both contextual factors (Jiang

and Conrath, 1997). The Jiang-Conrath similarity

takes into account the information content (IC) for

Teacher Educational Resources Recommendation in the COVID-19 Context

591

Algorithm 2: Teacher’s contextual factors weighting algorithm.

Input: C

work

: current working environment context of a teacher T

living

: current living environment context of a teacher T

sentimental

: current sentimental context of a teacher T

: history of current teacher T

FF : ﬁtness function

Output: W

f inal

={< f

>,. . . ,< f

>} : set of ﬁnal weights w

for each of the selected n

context factor

from the n initial context factors where n

< n.

begin

1: W

init

=selectFactors(C

work

, C

living

, C

sentimental

, H

);

2: W

f inal

=optimizeWeights (W

init

,FF);

3: return W

f inal

end

each selected factor (semantic concept) in the set

of contextual factors for C

and C

as in Eq. 2.

This similarity measure is not fulﬁlled until the

distance distance between any two selected factors

(semantic concepts) f

and f

is obtained using the

least common subsumer (LCS) as in Eq. 3. The

semantic similarity threshold thr

sim

is responsible for

omitting the least similar teachers from the ﬁnal list.

This teachers list is passed to be evaluated using the

SWRL rules.

Sim( f

, f

) =

D( f

, f

)

(2)

D( f

, f

) = IC( f

) + IC( f

)–(2 ∗ IC(LCS( f

, f

))

(3)

Afterwards, the list of teachers are sorted using

SWRL semantic reasoning rules in Table 2 to ensure

the correctness of each list element as shown in Al-

gorithm 5. A teacher in the list obtains a scoring

unit if a SWRL rule is satisﬁed. At the end, the to-

tal score of a teacher is compared against the SWRL

rules score threshold thr

swrl

. If the teacher does not

pass this condition, the list item will be eliminated.

The ﬁnal sorted list is returned to be used for provid-

ing resource recommendations.

Collaborative ﬁltering recommendation approach

is followed to provide educational resources’ recom-

mendations for the teacher T using the two similarity

measures that are calculated during the previous al-

gorithm steps: semantic similarity score and SWRL

rules similarity. Finally, the teacher is asked to inter-

act with these recommendations within MEMORAe

system, meanwhile, the activities and the sentimental

state are monitored.

4 DISCUSSION

In this approach, we enforce the sentimental state of

teachers during the COVID-19 pandemic context by

enhancing the process of ﬁnding proper educational

resources recommendations as introduced in Subsec-

tion 3.2. This enhancements are noticeable in two

main phases of the proposed approach: contextual

factors extraction in Algorithm 3 and the sorting of

the matching teachers list using SWRL rules in Al-

gorithm 5. The output results can be greatly affected

according to these new enhancements which is dis-

cussed during this section.

4.1 Contextual Factors Selection

For better evaluation of this enhancement’s impact

on the factors selection, we discuss it using an

actual example to compare resulting output with and

without the sentimental state condition in Algorithm

3, line 13.

Scenario 1: Let’s assume that Teacher 1 T

has

the following contextual factors for each context:

= {Age, Education, Language, TeacherLevel,

FieldofScience, . . . }

work

= {AreaType, Country, Population, . . . }

living

= {AreaType, Country, Population, . . . }

sentimental

= {MoodLevel, CurrentMood, Average-

Mood, . . . }

When the algorithm is executed without the

mentioned condition, it generates a vector of selected

and weighed contextual factors as follows:

W = {<Language,0.12>, <FieldofScience,0.35>,

<MoodLevel,0.16>, <WorkingPlace,0.33>,

<LivingLocation,0.14>}

EKM 2022 - 5th Special Session on Educational Knowledge Management

592

Algorithm 3: Variance-based teacher contextual factors selection algorithm.

Input: (C

resr

)

work

=(T ,F,rating,resr) : current working context of a teacher T and concerning the current re-

source resr with rating r, represented by context factors F.

resr

)

living

=(T ,F,rating,resr) : current living context of a teacher T and concerning the current resource resr

with rating r, represented by context factors F.

resr

)

sentimental

=(T ,F,rating,resr) : current sentimental context of a teacher T and concerning the current

resource resr with rating r, represented by context factors F.

: history of current teacher T .

thr

: given similarity threshold.

thr

: given weight threshold.

coe f

mood

: mood contextual factor enforcing coefﬁcient.

Output: W

init

={< f

init

>,. . . ,< f

init

>} : set of initial weights w

init

for each of the n

context factor f

begin

1: W

init

= {φ};

2: for all C

resr

∈ ((C

resr

)

work

,(C

resr

)

living

,(C

resr

)

sentimental

) do

3: for all f

∈ F do

4: nb

= 0;

5: Sim

total

= 0;

6: for all < resr

>∈ H

7: if (resr

== resr) then

8: similar(T

,T ) = calculateSimilarity(C

resr

);

9: if (similar(T

,T ) ≥ thr

) then

10: nb

+ +;

11: rating

= getRating(T

,resr);

12: similarTeachersList.add(rating

);

13: if ( f

== ‘mood’&& (getMood(T

) == ‘positive’ || getMood(T

) == ‘neutral’)) then

14: similar(T

,T ) *= coe f

mood

;

15: end if

16: Sim

total

+= similar(T

,T );

17: end if

18: end if

19: end for

20: sim

resr

Sim

total

;

21: sim

resr

var

= getVariance(similarTeachersList);

22: w

init

= average(sim

resr

,sim

resr

var

);

23: if (w

init

≥ thr

) then

24: W

init

= addFactor(< f

init

>);

25: end if

26: end for

27: end for

28: return W

init

;

end

The resulting contextual factors’ list shows that

the MoodLevel factor obtains low weight with

respect to other factors which accordingly affects the

ﬁnal list of recommendations.

However, when we re-run the same algorithm

with the mentioned condition, we obtain the follow-

ing vector:

W = {<Language,0.1>, <FieldofScience,0.35>,

<MoodLevel,0.4>, <WorkingPlace,0.28>,

<LivingLocation,0.05>}

Therefore, the difference between the weights

in both cases, illustrates the noticeable effect of

the enforcing coefﬁcient on the resulting vector of

contextual factors.

4.2 SWRL Rules Utilization

Through this subsection, we illustrate the effect of en-

forcing various measures using semantic SWRL rules

for a better context representation in this pandemic

period. The experience level (Rule 1), sentimental

Teacher Educational Resources Recommendation in the COVID-19 Context

593

Algorithm 4: Teacher context matching algorithm.

Input: (C

) : selected contextual factors of current teacher T

W ={< f

>,. . . ,< f

>} : set of ﬁnal weights w

for each of context factor f

thr

sim

: given similarity threshold.

Output: MatchingTeacherList

={< T

,Sim

>,. . . ,< T

,Sim

>} : list of all matching teachers to current

teacher T

begin

1: MatchingTeacherList= {φ};

2: for all C

∈ C

all

3: Sim

= 0;

4: for all f

∈ C

5: distance = IC( f

) + IC( f

) – ( 2 * IC(LCS( f

, f

)) );

6: sim( f

, f

) =

distance

;

7: Sim

+= sim( f

, f

) * w

;

8: end for

9: if Sim

>= thr

im then

10: MatchingTeacherList.add(< T

,Sim

>);

11: end if

12: end for

13: MatchingTeacherList.sort();

14: return MatchingTeacherList;

end

Algorithm 5: SWRL rules scoring algorithm.

Input: T

: current teacher

MatchingTeacherList : list of all matching teach-

ers to current teacher T

SW RL

={r

,. . . ,r

} : set of SWRL rules

thr

swrl

: given SWRL rules score threshold

Output: sortedList : sorted list of matching teachers

to current teacher T

with respect to the obtained

SWRL rules score

begin

1: MatchingTeacherList= {φ};

2: for all T

∈ MatchingTeacherList do

3: score

= 0;

4: for all r

∈ R

SW RL

5: result = applySWRL(r

);

6: if T

∈ result then

7: score

++;

8: end if

9: end for

10: if score

≥ thr

swrl

then

11: MatchingTeacherList[T

,score] = score

;

12: else

13: MatchingTeacherList.delete(T

);

14: end if

15: end for

16: sortedList = MatchingTeacherList.sortByScore();

17: return sortedList;

end

context of a teacher (Rule 2), working context (Rule

6,7), living context (Rule 4,5), and ﬁeld of experience

(ﬁeld of science) (Rule 8), spoken languages (Rule 3)

are the enforced factors in the SWRL rules scoring.

The SWRL rules is utilized to enhance the matching

similarity scores for the mentioned factors. Scenario

2 illustrates the impact of SWRL rules in enforcing

the desired factors.

Scenario 2: Assuming that teacher T

has a negative

mood while interacting with an educational resource,

it is found that teachers T

and T

are possible

matches for this teacher as shown in Fig. 4. The

extraction and weighting algorithm produces a list

of matching teachers: T

and T

with 0.76 and 0.62

respective scores. Therefore, the algorithm uses the

similarity between T

and T

to generate the list of

resources’ recommendation. The enforcement of

positive mood as a contextual factor, using the SWRL

rules, increases the similarity score of T

by 0.34

which means this teacher would not be included into

the list of matching teachers.

W = {<Lang,0.1>, <FieldofScience,0.35>,

<MoodLevel,0.4>, <WorkingPlace,0.28>,

<LivingLocation,0.05>}

= {{Lang1,Lang2}, FieldofScience1, Nega-

tiveMood, SmallCity, SmallCity}

= {{ Lang1, Lang2, Lang3}, FieldofScience1,

PositiveMood, LargeCity, SmallCity}

= {{Lang2,Lang3}, FieldofScience1, Negative-

EKM 2022 - 5th Special Session on Educational Knowledge Management

594

Mood, SmallCity, SmallCity}

Sim(1,2) = (1 ∗ 0.1 + 1 ∗ 0.35 + 1 ∗ 0.4 + 0 ∗

0.28 + 1 ∗ 0.05)/1.18 = 0.76

Sim(1,3) = (0.5 ∗ 0.1 + 1 ∗ 0.35 + 0 ∗ 0.4 + 1 ∗ 0.28 +

1 ∗ 0.05)/1.18 = 0.62

Accordingly, we can say that introducing these

SWRL rules provides resources that generate posi-

tive mood with other teachers. These SWRL rules

can be replaced with other roles or modiﬁed to fol-

low different approaches. For instance, if we replace

Rule 8 with another rule that ensures the exact match

between the teachers’ ﬁeld of science, the order of

the matching teachers’ list will be executed differ-

ently and therefore, the output recommendations will

be impacted. Moreover, we can increase the number

of rules to achieve more adjustment to the matching

list and accordingly, the ﬁnal provided recommenda-

tions, which provides us with a ﬂexible algorithm that

can achieve different levels of adaptation according to

our research motivation.

5 CONCLUSION

This paper endorses the usage of context-aware ap-

proach to recommend educational resources to teach-

ers during this COVID-19 era. Also, it introduces

the sentimental state enforcement during the con-

text matching phase which provides the teacher with

positive-mood recommendations. The proposed ap-

proach introduces new methodology for selecting and

sorting recommendations to generate the ﬁnal list us-

ing SWRL semantic reasoning rules. This work is

considered as a ﬁrst step to explore the context-aware

recommender approach and highlight the sentimental

state importance during these difﬁcult times. Further

assessments are needed to validate this proposal along

with real-time experiments.

This work is considered as a ﬁrst step to explore

the context-aware recommender approach and high-

light the sentimental state importance during these

difﬁcult times. Further assessments are needed to val-

idate this proposal along with real-time experiments.

In addition, it can be strengthened through exploring

different approaches. Using this approach, a dataset

can be created to facilitate the research in the teacher’s

sentimental state and its relation to the other contexts

in which the teacher coexists. This suggestion can

lead to a possible computational overhead avoidance.

Other approaches can be investigated with the factor’s

weight optimization using deep learning techniques.

Environment

TCO

is-a

LivingEnvironment:

Small City

TCO

is-a

WorkingEnvironment:

Small City

TCO

tco:evolvesIn

tco: worksIn

EducationalInstitution:

Higher-Education Institute

TCO

tco:livesIn

owl:equivalentClass

mcc:hasUserAccount

Person:

Teacher 1

TCO

Person:

Teacher 1

MCC

mcc:hasActor

UserAccount

MCC

InteractionActivity:

View resource 257

MCC

mdo:resultsIn

Mood:

Negative

MDO

Teacher:

Novice

TCO

is-a

language:

Lang.1, Lang.2

tco:hasLanguage

FieldofScience:

FieldofScience1

tco:hasScience

Environment

TCO

is-a

LivingEnvironment:

Small City

TCO

is-a

WorkingEnvironment:

Large City

TCO

tco:evolvesIn

tco: worksIn

EducationalInstitution:

Higher-Education Institute

TCO

tco:livesIn

owl:equivalentClass

mcc:hasUserAccount

Person:

Teacher 2

TCO

Person:

Teacher 2

MCC

mcc:hasActor

UserAccount

MCC

InteractionActivity:

View resource 257

MCC

mdo:resultsIn

Mood:

Positive

MDO

Teacher:

Intermediate

TCO

is-a

language:

Lang.1, Lang.2, Lang.3

tco:hasLanguage

FieldofScience:

FieldofScience1

tco:hasScience

Environment

TCO

is-a

LivingEnvironment:

Small City

TCO

is-a

WorkingEnvironment:

Small City

TCO

tco:evolvesIn

tco: worksIn

EducationalInstitution:

Higher-Education Institute

TCO

tco:livesIn

owl:equivalentClass

mcc:hasUserAccount

Person:

Teacher 3

TCO

Person:

Teacher 3

MCC

mcc:hasActor

UserAccount

MCC

InteractionActivity:

View resource 257

MCC

mdo:resultsIn

Mood:

Negative

MDO

Teacher:

Expert

TCO

is-a

language:

Lang.1, Lang.2

tco:hasLanguage

FieldofScience:

FieldofScience1

tco:hasScience

(1)

(2)

(3)

Figure 4: Partial Abox instance of the ontology representa-

tion of teachers 1, 2, and 3.

Teacher Educational Resources Recommendation in the COVID-19 Context

595

Table 2: SWRL rules for peer’s list sorting.

# SWRL Rule

1 tco:teacher(?t)ˆtco:hasYearsOfExperience(?t,?ex)ˆtco:teacher(?ts)ˆtco:

hasYearsOfExperience(?ts,?exs)ˆswrlb:greaterthan(?exs,?ex) =⇒ sqwrl:select(?ts)

2 mcc:InteractionActivity(?e)ˆmcc:hasActor(?e,?acc)ˆmdo:resultsIn(?e,?m)ˆmdo:mood(?m)

ˆmdo:hasValue(?m,?v)ˆmcc:hasActor(?e,?accs)ˆmdo:resultsIn(?e,?ms)ˆmdo:mood(?ms)ˆmdo:

hasValue(?ms,?vs)ˆswrlb:greaterthan(?vs,?v) =⇒ sqwrl:select(?ts)

3 tco:teacher(?t)ˆtco:hasLanguage(?t,?lan)ˆtco:teacher(?ts)ˆtco:hasLanguage(?ts,?lans)ˆswrlb:

contains(?lan,lans) =⇒ sqwrl:select(?ts)

4 tco:teacher(?t)ˆtco:livesIn(?t,?livenv)ˆtco:LivingEnvironment(?livenv)ˆtco:is-a(?livenv,?env)

ˆtco:environment(?env)ˆtco:hasType(?env,?envtype)ˆtco:teacher(?ts)ˆtco:livesIn(?ts,?livenvs)

ˆtco:LivingEnvironment(?livenvs)ˆtco:is-a(?livenvs,?envs)ˆtco:environment(?envs)ˆtco:

hasType(?envs,?envtypes)ˆswrlb:equal(?envtype,?envtypes) =⇒ sqwrl:select(?ts)

5 tco:teacher(?t)ˆtco:livesIn(?t,?livenv)ˆtco:LivingEnvironment(?livenv)ˆtco:is-a(?livenv,?env)

ˆtco:environment(?env)ˆtco:hasCountry(?env,?coun)ˆowl:country(?coun)ˆtco:teacher(?ts)ˆtco:

livesIn(?ts,?livenvs)ˆtco:LivingEnvironment(?livenvs)ˆtco:is-a(?livenvs,?envs)ˆtco:

environment(?envs)ˆtco:hasCountry(?envs,?couns)ˆowl:country(?couns)ˆswrlb:equal(?coun,

?couns) =⇒ sqwrl:select(?ts)

6 tco:teacher(?t)ˆtco:worksIn(?t,?inst)ˆtco:EducationalInstitution(?inst)ˆtco:evolvesIn(?inst,

?workenv)ˆtco:WorkingEnvironment(?worknv)ˆtco:is-a(?workenv,?env)ˆtco:

environment(?env)ˆtco:hasType(?env,?envtype)ˆtco:teacher(?ts)ˆtco:worksIn(?ts,

?insts)ˆtco:EducationalInstitution(?insts)ˆtco:evolvesIn(?insts,?workenvs)ˆtco:

WorkingEnvironment(?worknvs)ˆtco:is-a(?workenvs,?envs)ˆtco:environment(?envs)ˆtco:

hasType(?envs,?envtypes)ˆswrlb:equal(?envtype,?envtypes) =⇒ sqwrl:select(?ts)

7 tco:teacher(?t)ˆtco:worksIn(?t,?inst)ˆtco:EducationalInstitution(?inst)ˆtco:

hasEducationLevel(?inst,?edulvl)ˆdcterms:EducationLevel(?edulvl)ˆtco:teacher(?ts)ˆtco:

worksIn(?ts,?insts)ˆtco:EducationalInstitution(?insts)ˆtco:hasEducationLevel(?insts,?edulvls)

ˆdcterms:EducationLevel(?edulvls)ˆswrlb:equal(?edulvl,?edulvls) =⇒ sqwrl:select(?ts)

8 tco:teacher(?t)ˆtco:hasScience(?t,?sci)ˆmodsci:Science(?sci)ˆtco:teacher(?ts)ˆtco:

hasScience(?ts,?scis)ˆmodsci:Science(?scis)ˆswrlb:equal(?sci,?scis) =⇒ sqwrl:select(?ts)

– tco:teacher(?t)ˆtco:hasYearsOfExperience(?t,?ex)ˆsqrlb:lessthan(?ex,5) =⇒ tco:

noviceTeacher(?nt)ˆrdf:is-a(?t,?nt)

– tco:teacher(?t)ˆtco:hasYearsOfExperience(?t,?ex)ˆsqrlb:lessthan(?ex,10)ˆsqrlb:greaterthan(?ex,

5) =⇒ tco:intermediateTeacher(?it)ˆrdf:is-a(?t,?it)

– tco:teacher(?t)ˆtco:hasYearsOfExperience(?t,?ex)ˆsqrlb:greaterthan(?ex,10) =⇒ tco:

expertTeacher(?xt)ˆrdf:is-a(?t,?xt)

EKM 2022 - 5th Special Session on Educational Knowledge Management

596

REFERENCES

Abel, M.-H., Benayache, A., Lenne, D., and Moulin, C.

(2007). E-memorae: a content-oriented environment

for e-learning. In E-Learning Networked Environ-

ments and Architectures, pages 186–205. Springer.

Akour, A., Ala’a, B., Barakat, M., Kanj, R., Fakhouri,

H. N., Malkawi, A., and Musleh, G. (2020). The

impact of the covid-19 pandemic and emergency dis-

tance teaching on the psychological status of univer-

sity teachers: a cross-sectional study in jordan. The

American journal of tropical medicine and hygiene,

103(6):2391.

Andres, F., Blancan, B., Foutouhi, F., Ghinea, G., and Leite,

M. C. (2021). Overview of the 4th decor workshop

2021. In 37th IEEE International Conference on Data

Engineering.

Bishay, A. (1996). Teacher motivation and job satisfaction:

A study employing the experience sampling method.

Journal of undergraduate Sciences, 3(3):147–155.

Cobos, C., Rodriguez, O., Rivera, J., Betancourt, J., Men-

doza, M., Le

on, E., and Herrera-Viedma, E. (2013).

A hybrid system of pedagogical pattern recommenda-

tions based on singular value decomposition and vari-

able data attributes. Information Processing & Man-

agement, 49(3):607–625.

Cooper, K. and Olson, M. R. (2020). The multiple ‘i’s’ of

teacher identity. In Changing research and practice,

pages 78–89. Routledge.

De Meo, P., Messina, F., Rosaci, D., and Sarn

e, G. M.

(2017). Combining trust and skills evaluation to form

e-learning classes in online social networks. Informa-

tion Sciences, 405:107–122.

Flores, M. A. and Day, C. (2006). Contexts which shape and

reshape new teachers’ identities: A multi-perspective

study. Teaching and teacher education, 22(2):219–

232.

Frenzel, A. C. (2014). Teacher emotions.

Friedman, I. A. (2000). Burnout in teachers: Shat-

tered dreams of impeccable professional performance.

Journal of clinical psychology, 56(5):595–606.

Herman, K. C., Hickmon-Rosa, J., and Reinke, W. M.

(2018). Empirically derived proﬁles of teacher stress,

burnout, self-efﬁcacy, and coping and associated stu-

dent outcomes. Journal of Positive Behavior Interven-

tions, 20(2):90–100.

Heutte, J., Fenouillet, F., Kaplan, J., Martin-Krumm, C.,

and Bachelet, R. (2016). The eduﬂow model: A con-

tribution toward the study of optimal learning environ-

ments. In Flow experience, pages 127–143. Springer.

Imran, A. S., Muhammad, K., Fayyaz, N., Sajjad, M., et al.

(2021). A systematic mapping review on mooc rec-

ommender systems. IEEE Access.

Jain, S., Lall, M., and Singh, A. (2021). Teachers’ voices

on the impact of covid-19 on school education: Are

ed-tech companies really the panacea? Contemporary

Education Dialogue, 18(1):58–89.

Jenaro, C., Flores, N., and Arias, B. (2007). Burnout and

coping in human service practitioners. Professional

psychology: research and practice, 38(1):80.

Jiang, J. J. and Conrath, D. W. (1997). Semantic similarity

based on corpus statistics and lexical taxonomy. arXiv

preprint cmp-lg/9709008.

Kla

snja-Mili

cevi

c, A., Vesin, B., and Ivanovi

c, M. (2018).

Social tagging strategy for enhancing e-learning expe-

rience. Computers & Education, 118:166–181.

Lachowska, B. H., Szpringer, M., Lachowski, S., and Pory-

dzaj, K. (2018). Conﬂict and facilitation between fam-

ily and occupational roles versus work-related stress

in the teachers’ group. Medical Studies/Studia Medy-

czne, 34(1):41–56.

Lizana, P. A., Vega-Fernadez, G., Gomez-Bruton, A., Ley-

ton, B., and Lera, L. (2021). Impact of the covid-19

pandemic on teacher quality of life: A longitudinal

study from before and during the health crisis. Inter-

national Journal of Environmental Research and Pub-

lic Health, 18(7):3764.

Marini, F. and Walczak, B. (2015). Particle swarm opti-

mization (pso). a tutorial. Chemometrics and Intelli-

gent Laboratory Systems, 149:153–165.

McIntyre, T. M., McIntyre, S. E., Barr, C. D., Francis, D. J.,

and Durand, A. C. (2017). Towards a dynamic inte-

grative theory of educator stress. In Educator Stress,

pages 261–289. Springer.

erida-L

opez, S. and Extremera, N. (2017). Emotional

intelligence and teacher burnout: A systematic re-

view. International Journal of Educational Research,

85:121–130.

Moore, P., Zhao, Z., and Pham, H. V. (2019). Towards

cloud-based personalised student-centric context-

aware e-learning pedagogic systems. In Conference

on Complex, Intelligent, and Software Intensive Sys-

tems, pages 331–342. Springer.

Mukhtar, K., Javed, K., Arooj, M., and Sethi, A. (2020).

Advantages, limitations and recommendations for on-

line learning during covid-19 pandemic era. Pakistan

journal of medical sciences, 36(COVID19-S4):S27.

Nashed, N. N., Lahoud, C., and Abel, M.-H. (2022). Edu-

cational resources recommender system for teachers:

Why and how? In Advances in Deep Learning, Artiﬁ-

cial Intelligence and Robotics, pages 71–80. Springer.

Nashed, N. N., Lahoud, C., Abel, M.-H., Andr

es, F., and

Blancan, B. (2021). Mood detection ontology integra-

tion with teacher context. In 2021 20th IEEE Interna-

tional Conference on Machine Learning and Applica-

tions (ICMLA), pages 1710–1715. IEEE.

Nilashi, M., Asadi, S., Abumalloh, R. A., Samad, S., and

Ibrahim, O. (2020). Intelligent recommender systems

in the covid-19 outbreak: the case of wearable health-

care devices. Journal of Soft Computing and Decision

Support Systems, 7(4):8–12.

Oberle, E., Gist, A., Cooray, M. S., and Pinto, J. B.

(2020). Do students notice stress in teachers? asso-

ciations between classroom teacher burnout and stu-

dents’ perceptions of teacher social–emotional com-

petence. Psychology in the Schools, 57(11):1741–

1756.

Peimani, N. and Kamalipour, H. (2021). Online educa-

tion and the covid-19 outbreak: A case study of on-

Teacher Educational Resources Recommendation in the COVID-19 Context

597

line teaching during lockdown. Education Sciences,

11(2):72.

Perifanou, M., Economides, A., and Tzaﬁlkou, K. (2021).

Teachers’ digital skills readiness during covid-19 pan-

demic.

Riedl, R., Kindermann, H., Auinger, A., and Javor, A.

(2012). Technostress from a neurobiological perspec-

tive. Business & Information Systems Engineering,

4(2):61–69.

Scherer, R., Howard, S. K., Tondeur, J., and Siddiq, F.

(2021). Proﬁling teachers’ readiness for online teach-

ing and learning in higher education: Who’s ready?

Computers in human behavior, 118:106675.

Skaalvik, E. M. and Skaalvik, S. (2020). Teacher burnout:

relations between dimensions of burnout, perceived

school context, job satisfaction and motivation for

teaching. a longitudinal study. Teachers and Teach-

ing, 26(7-8):602–616.

Thulin, E., Vilhelmson, B., and Johansson, M. (2019). New

telework, time pressure, and time use control in every-

day life. Sustainability, 11(11):3067.

Zhang, Q., Lu, J., and Zhang, G. (2021). Recommender

systems in e-learning. Journal of Smart Environments

and Green Computing, 1(2):76–89.

EKM 2022 - 5th Special Session on Educational Knowledge Management

598