Design Recommendations for Intelligent Tutoring Systems Volume 1

Abstract

The aim of this research was to develop a learning recommendation component in an intelligent tutoring system (ITS) that dynamically predicts and adapts to a learner's style. In order to develop a proper ITS, we present an improved knowledge base supporting adaptive learning, which can be achieved by a suitable knowledge construction. This process is illustrated by implementing a web-based online tutor system. In addition, our knowledge structure provides adaptive presentation and personalized learning with the proposed adaptive algorithm, to retrieve content according to individual learner characteristics. To demonstrate the proposed adaptive algorithm, pre-test and post-test were used to evaluate suggestion accuracy of the course in a class for adapting to a learner's style. In addition, pre- and post-testing were also used with students in a real teaching/learning environment to evaluate the performance of the proposed model. The results show that the proposed system can be used to help students or learners achieve improved learning.

Keywords

• Computer Science
• Learning recommendation
• Formal concept analysis
• Intelligent tutoring system
• Adaptive learning

1. Introduction

E-learning systems are important in online education, using information and communication technology to help the teacher and improve the self-study of a learner. Computer-aided instruction (CAI) [

1

• Goyal M.
• Yadav D.
• Choubey A.

E-learning: current state of art and future prospects.

• Google Scholar

,

] is one model that applies computers to education to support the self-study of a learner. However, CAI lacks the ability to adapt according to a student's learning style. Hence, intelligent tutoring systems (ITS) have been developed for adaptive educational systems that employ artificial intelligence (AI) techniques to provide individualized instruction [

3

• Klašnja-Milićević A.
• Vesin B.
• Ivanović Mirjana
• Budimac Z.

E-learning personalization based on hybrid recommendation strategy and learning style identification.

• Crossref
• Scopus (391)
• Google Scholar

,

].

ITSs are adaptive systems, which use intelligent technologies to personalize learning according to individual student characteristics, such as knowledge of the subject, mood and emotion, and learning style [

]. Many researchers have developed ITSs with various techniques to adapt the style of tutoring to match a student's learning style and enhance learning, such as Protus [

[3]

• Klašnja-Milićević A.
• Vesin B.
• Ivanović Mirjana
• Budimac Z.

E-learning personalization based on hybrid recommendation strategy and learning style identification.

• Crossref
• Scopus (391)
• Google Scholar

], Oscar [

], ZOSMAT [

], etc. The traditional ITS architecture involves an expert model, instructional model, learner model, and user interface [

,

7

• Kuo R.
• Hsu C.K.
• Chang M.
• Heh J.

A personalized webpage reconstructor based on concept lattice and association rules.

• Google Scholar

]. To improve the value of ITSs, these models will use a knowledge structure [

,

] to personalize learning [

3

• Klašnja-Milićević A.
• Vesin B.
• Ivanović Mirjana
• Budimac Z.

E-learning personalization based on hybrid recommendation strategy and learning style identification.

• Crossref
• Scopus (391)
• Google Scholar

,

,

11

• Taran T.
• Sirota S.

Knowledge learning technology for intelligent tutoring systems.

• Google Scholar

]; and adaptive learning [

,

] to improve the effectiveness of a learner's experience. Thus, approaches to construct knowledge base and adaptive learning are key components in a successful ITS.

In order to develop a proper ITS, we require a knowledge base supporting adaptive learning, which can be achieved by a suitable knowledge construction, i.e., a knowledge representation with knowledge acquisition to personalize learning according to an individual learner's characteristics. For this reason, this work provides formal concept analysis (FCA) [

,

13

• Johnson W.L.
• Valente A.

Tactical language and culture training systems: using artificial intelligence to teach foreign languages and cultures.

• Google Scholar

,

14

• van Seters J.R.
• Ossevoort M.A.
• Tramper J.
• Goedhart M.

The influence of student characteristics on the use of adaptive e-learning material.

• Crossref
• Scopus (68)
• Google Scholar

,

] for building the knowledge structure in ITS to support knowledge acquisition and adaptive learning. FCA is one approach for grouping documents into a hierarchical form that supports browsing. It automatically provides generalization and specialization relationships among the formal concepts of a document structure in a concept lattice [

]. Thus, this work applied FCA to adaptive learning for learners. Moreover, this method provides the content of related courses or group of documents that enable recommendations for a learning style. Adaptive learning consists of adaptive presentation and adaptive navigation [

,

]. Adaptive presentation [

,

,

14

• van Seters J.R.
• Ossevoort M.A.
• Tramper J.
• Goedhart M.

The influence of student characteristics on the use of adaptive e-learning material.

• Crossref
• Scopus (68)
• Google Scholar

] refers to the education material while adaptive navigation [

,

,

,

18

• Wille R.

Formal concept analysis as mathematical theory of concepts and concept hierarchies.

• Google Scholar

] refers to manipulation of the links and structure. This research focuses on adaptive presentation to give personalized advice on content to the learner by using FCA approach.

This article is organized as follows. In Section 2, we describe related works. In Section 3, we present our proposed learning recommendation component for ITS. Section4 shows an example and discusses how our proposed framework can be beneficial. Finally, Section 5 concludes and envisions future works.

2. Related works

This section provides background and related work on ITS and FCA in subsections 2.1 and 2.2, respectively. Related works are reviewed for adaptive learning in ITS in subsection 2.3. In particular, the current research and proposed methods are discussed.

2.1 Intelligent tutoring systems

An Intelligent Tutoring System (ITS) [

,

,

,

7

• Kuo R.
• Hsu C.K.
• Chang M.
• Heh J.

A personalized webpage reconstructor based on concept lattice and association rules.

• Google Scholar

,

19

• Mohamed H.
• Bensebaa T.
• Trigano P.

Developing adaptive intelligent tutoring system based on item response theory and metrics.

• Google Scholar

,

,

21

• Zhu G.
• Zhang Z.

Knowledge unit discovery for programming tutoring based on formal concept analysis.

• Google Scholar

] involves presenting learning materials to a learner as guided by artificial intelligence (AI) techniques, so as to improve the effectiveness of a learner's experience. Generally, the classical ITS architecture is considered to have four major components: expert model, instructional model, learner model, and user interface as shown in Fig. 1.

In Fig. 1, the first model is expert knowledge. The expert knowledge model [

,

,

7

• Kuo R.
• Hsu C.K.
• Chang M.
• Heh J.

A personalized webpage reconstructor based on concept lattice and association rules.

• Google Scholar

] comprises the facts and rules of the particular domain to be conveyed to the learner, i.e. the knowledge of the experts. Knowledge elicitation and codification can be a very time-consuming task, especially for a complex domain with an enormous amount of knowledge and interrelationships of that knowledge. Thus, investigating how to encode knowledge and how to represent it in an ITS remains the central issue of creating an expert knowledge model. Also, in its function as a standard, the expert knowledge model can be used to assess the student's overall progress. To achieve this requires establishing some criteria for comparing levels of knowledge. This type of comparison is possible only if the knowledge has been explicitly represented. Hence, ITSs differ considerably from traditional CAI programs in that the knowledge in the latter is implicitly represented within its code.

Instructional model [

,

21

• Zhu G.
• Zhang Z.

Knowledge unit discovery for programming tutoring based on formal concept analysis.

• Google Scholar

,

] refers to the teaching strategy or the pedagogic model. This model contains knowledge for making decisions about instructional tactics. It relies on the diagnostic processes of the learner model for making decisions about what information to present, and when and how, to a leaner.

Learner model [

,

,

23

• Kularbphettong K.

An adaptive web-based intelligent tutoring using mastery learning and logistic regression techniques.

• Google Scholar

] is often viewed as a subset of the expert knowledge model, which changes over the course of tutoring. ITS is able to adapt to a specific student during the teaching process. The learner model can be evaluated in regard to several criteria of usefulness. These criteria include the following: (1) Learner model fit to data refers to how well the learner model can be used to simulate the quantitative and/or qualitative patterns of learning in real students [

[1]

• Goyal M.
• Yadav D.
• Choubey A.

E-learning: current state of art and future prospects.

• Google Scholar

], (2) Ease of understanding, (3) Generality/flexibility, (4) Cost of creation, (5) Multiple choices to pedagogical decisions at any level, (6) Time scale refers to the overall longevity of the learner model, and (7) Learning gains in practice. Finally, user interface module is the communicating component of the ITS, which controls interaction between the learner and the system. In addition to the conventional human-computer interface features, some recent systems have had natural language interaction [

,

], speech recognition [

,

27

• Cheng S.Y.
• Lin C.S.
• Chen H.
• Heh J.

Learning and diagnosis of individual and class conceptual perspectives: an intelligent systems approach using clustering techniques.

• Crossref
• Scopus (32)
• Google Scholar

], and the sensing of student emotions [

27

• Cheng S.Y.
• Lin C.S.
• Chen H.
• Heh J.

Learning and diagnosis of individual and class conceptual perspectives: an intelligent systems approach using clustering techniques.

• Crossref
• Scopus (32)
• Google Scholar

,

].

Adaptive hypermedia [

,

7

• Kuo R.
• Hsu C.K.
• Chang M.
• Heh J.

A personalized webpage reconstructor based on concept lattice and association rules.

• Google Scholar

,

] and ITS are used in E-learning systems to drive the connection and to personalize instruction on the basis of adaptation to the learner's learning style. An ITS needs to be designed and implemented to support modification of lecture content, the decision rules and the fact base of the expert model, and the methods to measure performance of learning. There are two forms of adaptive learning, i.e. adaptive presentation and adaptive navigation, which are mentioned in subsection 2.3.

ITSs are adaptive systems that use intelligent technology to personalize learning according to an individual learner's characteristics [

3

• Klašnja-Milićević A.
• Vesin B.
• Ivanović Mirjana
• Budimac Z.

E-learning personalization based on hybrid recommendation strategy and learning style identification.

• Crossref
• Scopus (391)
• Google Scholar

,

,

11

• Taran T.
• Sirota S.

Knowledge learning technology for intelligent tutoring systems.

• Google Scholar

]. There are three main approaches to intelligent tutoring: (1) Curriculum sequencing that aims to improve the learning experience, (2) intelligent solution analysis, focusing on giving students detailed feedback on incomplete or erroneous solutions, helping them learn from their mistakes, and (3) Problem solving support, involving a constructivist approach, for students constructing their own knowledge, and thus encouraging a deeper understand of the topics. Furthermore, various ITSs were built for interactive learning with the learners surveyed [

,

30

• Dašić P.
• Dašić J.
• Crvenković B.
• Šerifi V.

A review of intelligent tutoring systems in e-learning.

• Google Scholar

]. Table 1 summarizes reviewed prior studies of interest. It shows that mostly the previous works have applied machine learning with an expert model based on rule-based reasoning.

Table 1 Summary of relevant prior studies.

Study	Methodology	Domain	Summary
Harley et al. (2015) [31] Harley J.M. Bouchet F. Hussain M.S. Azevedo R. Calvo R. A multi-componential analysis of emotions during complex learning with an intelligent multi-agent system. Crossref Scopus (99) Google Scholar	Emotional measurement methods	Human Circulatory	Providing a methodological description of how to measure and synchronize emotion data obtained from learners interacting with multi-agent computerized learning environment.
Grawemeyer et al. (2016) [32] Grawemeyer B. Mavrikis M. Holmes W. Gutierrez-Santos S. Wiedmann M. Rummel N. Affecting off-task behaviour: how affect-aware feedback can improve student learning. Crossref Scopus (19) Google Scholar	Application from iTalk2Learn that is a learning platform for children aged 8-12 years old who are learning fractions.	Mathematics	Using two Bayesian networks to predict a student's affective state and adapting the type of feedback and the presentation of the feedback according to their affective states. The results point to the potential and positive impact of affect- aware intelligent support.
Grivokostopoulou et al. (2017) [33] Grivokostopoulou F. Perikos I. Hatzilygeroudis I. An educational system for learning search algorithms and automatically assessing student performance. Crossref Scopus (27) Google Scholar	Application for interactive examples and exercises related to search algorithms.	Artificial intelligence curriculum	Using Bsearch algorithms to offer theory descriptions, interactive examples and exercises related to search algorithms in Artificial intelligence tutoring system. The results show that visualized animations and interactivity are two crucial factors that contribute in better learning.
Kose and Arslan (2017) [34] Kose U. Arslan A. Optimization of self-learning in computer engineering courses: an intelligent software system supported by artificial neural network and vortex optimization algorithm. Crossref Scopus (23) Google Scholar	Evaluate students' intelligence levels according to the Theory of Multiple Intelligences	Computer engineering	The system is based on a hybrid artificial intelligence approach formed by an Artificial Neural Network, and an optimization algorithm called Vortex Optimization Algorithm (VOA). It has been tested with students giving positive results on optimizing for improved self-learning.
Khachatryan et al. (2014) [35] Khachatryan G.A. Romashov A.V. Khachatryan A.R. Gaudino S.J. Khachatryan J.M. Guarian K.R. Yufa N.V. Reasoning mind genie 2: an intelligent tutoring system as a vehicle for international transfer of instructional methods in mathematics. Crossref Scopus (30) Google Scholar	Application for interactive	Mathematics	Proposes the design of intelligent tutoring systems intended for cross-cultural transfer of curriculum and instructional methods. The results show that it is possible to use systems to implement one country's curriculum and instructional practices in another country in a way that has high acceptance.
Weragama and Reye (2014) [36] Weragama D. Reye J. Analysing student programs in the php intelligent tutoring system. Crossref Scopus (15) Google Scholar	Using Artificial Intelligence (AI), first-order predicate logic, classical, hierarchical planning to model the subject matter taught by the system.	Computer programming	Using PHP ITS to analyze students' programs, demonstrating ability to correctly analyze over 96 % of the solutions to exercises supplied by students.
Soh Khandaker and Jiang (2008) [37] Soh L.K. Khandaker N. Jiang H. I-minds: a multiagent system for intelligent computer-supported collaborative learning and classroom management. Google Scholar	Using synchronous learning and classroom management applications	Computer programming	Using a computer-supported collaborative learning (CSCL) infrastructure and environment for learners for supporting the instructor. Results show that it can be used in a real-time environment to support student cooperative learning activities.
Latham et al. (2012) [38] Latham A. Crockett K. McLean D. Edmonds B. A conversational intelligent tutoring system to automatically predict learning styles. Crossref Scopus (109) Google Scholar	learning style and personalizing the tutorial	Computer programming	Implicit modeling of the learning style during tutoring and personalizing the tutorial to boost confidence and improve the effectiveness of the learning experience.
Wang et al. (2015) [39] Wang D. Han H. Zhan Z. Xu J. Liu Q. Ren G. A problem solving oriented intelligent tutoring system to improve students' acquisition of basic computer skills. Crossref Scopus (31) Google Scholar	imitate the human teachers' actions with interactive learning environments	Basic computer skills	The concepts and methods of designing problem solving oriented ITS, which can imitate the human teachers' actions to implement one-to-one personalized teaching to some extent, for use as effective tool in training problem solving ability.
Keles and Gülcü (2009) [5] Keleş A. Ocak R. Keleş A. Gülcü A. Zosmat: Web-based intelligent tutoring system for teaching–learning process. Crossref Scopus (50) Google Scholar	Web-based application based on rule-base system.	Mathematics	Developed an intelligent tutoring system with attempts to imitate the behavior of an intelligent human tutor and a domain expert. The results show that this system is a powerful tool to improving the learning performance of a student.

• Open table in a new tab

However, the studies are limited by the fixed questions and content to suggest to users [

37

• Soh L.K.
• Khandaker N.
• Jiang H.

I-minds: a multiagent system for intelligent computer-supported collaborative learning and classroom management.

• Google Scholar

,

,

39

• Wang D.
• Han H.
• Zhan Z.
• Xu J.
• Liu Q.
• Ren G.

A problem solving oriented intelligent tutoring system to improve students' acquisition of basic computer skills.

• Crossref
• Scopus (31)
• Google Scholar

]. Fixed exercises are provided to learner with focus on learner model [

32

• Grawemeyer B.
• Mavrikis M.
• Holmes W.
• Gutierrez-Santos S.
• Wiedmann M.
• Rummel N.

Affecting off-task behaviour: how affect-aware feedback can improve student learning.

• Crossref
• Scopus (19)
• Google Scholar

,

34

• Kose U.
• Arslan A.

Optimization of self-learning in computer engineering courses: an intelligent software system supported by artificial neural network and vortex optimization algorithm.

• Crossref
• Scopus (23)
• Google Scholar

,

35

• Khachatryan G.A.
• Romashov A.V.
• Khachatryan A.R.
• Gaudino S.J.
• Khachatryan J.M.
• Guarian K.R.
• Yufa N.V.

Reasoning mind genie 2: an intelligent tutoring system as a vehicle for international transfer of instructional methods in mathematics.

• Crossref
• Scopus (30)
• Google Scholar

]. Interactive systems can be quite time consuming to develop [

[33]

• Grivokostopoulou F.
• Perikos I.
• Hatzilygeroudis I.

An educational system for learning search algorithms and automatically assessing student performance.

• Crossref
• Scopus (27)
• Google Scholar

]. The rules provided for application are not dynamic. These limitations are eliminated by using a dynamic knowledge base or an amendable knowledge base as provided by FCA. In this research, we focus on adaptive presentation as the proposed improved alternative type of knowledge base, by using FCA for ITS. We use intelligent solution analysis approach to improve self-study by students, evaluated by assessing changes from pre-test before study to next test time.

2.2 Formal concept analysis (FCA)

Formal concept analysis (FCA) [

,

13

• Johnson W.L.
• Valente A.

Tactical language and culture training systems: using artificial intelligence to teach foreign languages and cultures.

• Google Scholar

,

14

• van Seters J.R.
• Ossevoort M.A.
• Tramper J.
• Goedhart M.

The influence of student characteristics on the use of adaptive e-learning material.

• Crossref
• Scopus (68)
• Google Scholar

,

,

40

• Ganter B.
• Wille R.

Applied Lattice Theory: Formal Concept Analysis.

• Google Scholar

], is a mathematical approach to data analysis based on lattice theory. It is widely used in information science [

[13]

• Johnson W.L.
• Valente A.

Tactical language and culture training systems: using artificial intelligence to teach foreign languages and cultures.

• Google Scholar

] to describe attributes of information represented in a hierarchical model. FCA is not only a method for data analysis and knowledge representation, but also a formal formulation for concept formation and learning. It provides both generalization and specialization among concepts through the concept lattice [

14

• van Seters J.R.
• Ossevoort M.A.
• Tramper J.
• Goedhart M.

The influence of student characteristics on the use of adaptive e-learning material.

• Crossref
• Scopus (68)
• Google Scholar

,

]. Practically, FCA starts with a formal context, which contains values 0 or 1 in an information system. Below, we introduce basic definitions and ideas of FCA following reference [

].

Definition 1: A formal context $K:=(G,M,I)$ consists of two sets G and M and a relation I between G and M. The elements of G are called the objects and the elements of M are called the attributes of the context. In order to express that an object g is in a relation I with an attribute m, we write gIm or (g,m) ∈ I and read it as "the object g has the attribute m".

Another form of formal context is called a cross table. It can be used to identify groups of cases with shared attributes in binary relation format. In knowledge representation of this study, the objects represent each chapter of each class, and are classified by the topics in a course. Attributes are based on sets of keywords from all the content in the class. This study introduces the definition of an association between object and its attributes as follows.

Definition 2: For a set A ⊆ G of objects, A' is defined as follows

Correspondingly, for a set B ⊆ M of attributes, B' is defined as follows

Definition 3: A formal concept of the formal context (G, M, I) is a pair (A, B) with A ⊆ G, B ⊆ M, A'=B and B'=A. We call A the extent and B the intent of the formal concept (A, B). B(G, M, I) denotes the set of all formal concepts of the formal context (G, M, I).

Definition 4: If ( $A_1$, $B_1$) and ( $A_2$, $B_2$) are formal concepts of formal context K:= (G, M, I), ( $A_1$, $B_1$) is called a subconcept of ( $A_2$, $B_2$) (or ( $A_2$, $B_2$) is a superconcept of ( $A_1$, $B_1$)), provided that $A_1\subseteq A_2$ (or $B_2\subseteq B_1$) and this is denoted by ( $A_1,B_1$) ≤ ( $A_2,B_2$). The relation ≤ is called the hierarchical order (or simply order) of the formal concepts. The set of all formal concepts of (G, M, I) partially ordered in this way is denoted by B(G, M, I) and is called the concept lattice of the formal context (G, M, I).

To extract knowledge for advising a learner, the implications between keywords from content are used to identify chapters that the learner should study after taking a post-test. The implication provides a group of chapters identified by keywords. Mostly, the previous studies with FCA applied to ITS used it for adaptive navigation in an e-learning system [

,

,

7

• Kuo R.
• Hsu C.K.
• Chang M.
• Heh J.

A personalized webpage reconstructor based on concept lattice and association rules.

• Google Scholar

]. The authors applied FCA as a knowledge structure to webpages and their keywords [

,

]. These keywords relate to learner search domains. Next, these search keywords are checked in formal concepts for match with the knowledge structure, to receive webpage suggestions for the learner. Similarly, in [

], the authors demonstrated FCA to advice on topics in a Geometry course. In [

[18]

• Wille R.

Formal concept analysis as mathematical theory of concepts and concept hierarchies.

• Google Scholar

], FCA is applied to obtain association rules among keywords in webpages for Java language materials, to provide suitable content to learners.

The advantage of FCA in those prior studies in knowledge construction was based on it providing related keywords that can identify sets of data for adaptive navigation in an e-learning system. However, the adaptive presentation in the e-learning system should display educational materials according to a student's weaknesses, as indicated by their responses to questions based on earlier educational material. Thus, this study employs a concept lattice in FCA as a knowledge structure for adaptive presentation in ITS.

2.3 Adaptive learning in ITS

Adaptive learning [

,

,

,

41

• Dabolinš J.

Trends of the usage of adaptive learning in intelligent tutoring systems.

• Google Scholar

,

,

] is an educational method which uses computer as interactive teaching device. There are two forms, namely adaptive presentation and adaptive navigation. Adaptive presentation [

,

,

14

• van Seters J.R.
• Ossevoort M.A.
• Tramper J.
• Goedhart M.

The influence of student characteristics on the use of adaptive e-learning material.

• Crossref
• Scopus (68)
• Google Scholar

] refers to the education material display according to a student's weaknesses as indicated by their responses to questions based on earlier educational material. In other words, adaptive navigation [

,

,

,

18

• Wille R.

Formal concept analysis as mathematical theory of concepts and concept hierarchies.

• Google Scholar

] refers to the manipulation of links and structure. ITS can monitor the performance of a learner and personalize instruction on the basis of adaptation to the learner's learning style, current knowledge level, and appropriate teaching strategies in an E-learning system. Personalized learning [

3

• Klašnja-Milićević A.
• Vesin B.
• Ivanović Mirjana
• Budimac Z.

E-learning personalization based on hybrid recommendation strategy and learning style identification.

• Crossref
• Scopus (391)
• Google Scholar

,

] designs educational experiences that fit the need, goals, talents, and interests of the learner. Researchers have recently begun to investigate various techniques to help teachers in E-learning systems to assess learning styles [

,

]. Moreover, to adapt presentation and navigation according to learner, algorithms suggesting further learning can be found in prior literature [

,

14

• van Seters J.R.
• Ossevoort M.A.
• Tramper J.
• Goedhart M.

The influence of student characteristics on the use of adaptive e-learning material.

• Crossref
• Scopus (68)
• Google Scholar

,

,

].

In this research, we provide an improved knowledge base by using FCA to support adaptive presentation and personalized learning. This knowledge structure requires an adaptive algorithm to retrieve content according to an individual learner's characteristics. Thus, we also propose an algorithm to suggest courses and classes to each learner.

3. Research methodology

3.1 System architecture and design

In this section, we present our proposed ITS system in Fig. 2. Generally, the classical ITS architecture is considered to have four major components: expert model, instructional model, learner model, and user interface, which are described as follows.

The expert model comprises the facts and rules of the particular domain to be conveyed to the learner, i.e. the knowledge of the experts. Referring to Fig. 2, after teacher inputs the details of a course, that course will be generated using FCA to represent the knowledge. This knowledge is about the relationship between learner profiles and details of course to guide content for the learner. To extract knowledge for advising a learner, the implications between keywords from content are used to identify chapters that learners should study after they take a post-test. The implication provides a group of chapters identified by keywords.

Instructional model is referred to for the teaching strategy or the pedagogic model. This model contains knowledge for making decisions about instructional tactics. It relies on the diagnostic processes of the learner model for making decisions about what, when and how to present information to a learner. Instructional model in Fig. 2 is the education module. This process derived from the knowledge base shows content for recommended learning. After the learners study, they will test themselves. The score will identify the level of learning and is stored in their profile for future.

From Fig. 2, the learner profile will be recorded for use in adaptive learning according to course selection. The data in this learner profile is related to the knowledge base by using FCA to suggest content to learner in the next learning step. In this work, we created user interfaces for learner and teacher with natural language interaction. We will show an example in section 4.

3.2 Learning recommendation with adaptive algorithm

We apply FCA to collaborative recommendations for each learner. Fig. 3 illustrates an overview of knowledge acquisition and adaptive learning for learners. Firstly, FCA is applied to achieve a suitable knowledge construction. The course details that include contents, objectives, exercises, and tests, will be prepared and transformed into independent hierarchical structure with keyword acquisition.

The following example illustrates knowledge based acquisition of the expert model in our ITS system. Suppose the initial course details are as shown in Table 2. This formal context form is transformed into formal concept and concept lattice according to the Definition 2, Definition 4 as shown in Fig. 4. This figure represents knowledge base structure that shows the co-appearances and subconcept–superconcept relations in a concept lattice. This structure can provide a learning path to a learner. In Fig. 4, an example of a learning path is as follows:

Table 2 Document collection in formal context form.

	obj 1	obj 2	obj 3	k 1	k 2	k 3	k 4	k 5
Q1	1	0	0	1	0	1	1	0
Q2	1	1	0	0	1	0	0	1
Q3	0	1	1	1	0	1	0	1
Q4	1	0	1	0	1	0	1	0
Q5	0	1	0	0	1	0	0	0
Ex1	1	0	0	1	0	0	1	0
Ex2	1	0	1	1	0	1	0	1
Ex3	0	1	0	0	1	0	0	1

• Open table in a new tab

1. k3 → k1

2. k4 → obj1

3. obj3 and k5 → k1and k

4. k1 and k5 → obj3 and k3

5. k2 and k5 → obj2

From the above implication rules, the first and second rules imply that if the learners study topic k3 then they should study topic k1, and that if the learners study topic k4 then they will get to the first objective, respectively. These rules are used to recommend what to study to each learner.

Afterwards, the set of keywords is reformed into knowledge base structure by Algorithms 1 and 2 [

,

,

]. However, these algorithms in the previous work were proposed for a general case of FCA construction. Thus, this work proposes adapted algorithms for supporting ITS model. Next, when the learners study lessons and work through exercises and tests, they make mistakes in some questions, which informs about missed objectives. The keywords for mistaken exercises or tests will be matched with implication rules from knowledge base to identify relevant chapters or lessons. Consequently, our system will interact with the learners and recommend lessons to study again, and practice exercises and tests.

Algorithm 1 consists of two parts (i.e. extent and intent), where input data is a formal context

$(G,M,I)$

. G represents keywords and M represents objectives of each lesson. The first part, FindExtentX() is a function to generate a family of extents, while the rest is used to find the intent of the formal concept. The result is a formal concept that identifies groups of keywords sharing common objectives of each lesson. To extract knowledge for advising the learner, the implications between keywords from content are used to identify chapters that learners should study after they take a test. The implications provide groups of keywords identified by objectives. Thus, Algorithm 2 is provided to build a concept lattice to prepare related knowledge of content to the learner. Algorithm 3 finds implication rules to recommend to learners.

4. Implementation

4.1 System overview

The aim of this research was to develop learning recommendations in ITS that dynamically predict and adapt to a learner's style, by using smartphone and web applications. This work implemented a C programming language tutor for general C programming teachers and learners, called Tutor C Programming. The proposed system is shown as an overview in Fig. 5. This figure shows an example of application. The application is divided into two parts, for front-end user and for back-end user. The front-end user is served to learner and teacher (or admin) based on web application, while the back-end is a hidden module using FCA for expert model in ITS, mentioned in section 4.2. The system is designed for three user groups, namely learners, teachers, and administrators, with interface structure designs in Fig. 5.

Referring to Fig. 5, the user can register to the system. Next, the learner user can manage his/her profile to update data and study lessons by themselves, with content provided by teacher. They can take the exercises and tests, after which they are assessed to generate recommendations by system. Moreover, they can view the study history and results. The teacher can add content to his/her course. The proposed system supports many classes. The teacher can manage a lesson, the objectives of a lesson, exercises, and tests to the students. Moreover, the teacher can view the scores to assess improvements of his/her students. The final user type, administrators, can manage all information in the system.

The web-based application is implemented to manage information of learners, teachers, and courses to recommend to learner, shown in Fig. 6. This shows an example web page to illustrate information management for learners, teachers, and administrators (more detail can be viewed on the web: https://www.parasystem.org/LearnC/pages/login.php). The learner can study provided content or can take an examination to get learning recommendations that are described in the next subsection.

4.2 System with recommendations implemented

This section presents the implemented learning recommendation for learners. The learner can study from provided content or can take an examination to get the learning recommendations. The examination provided is one of two types, i.e. multiple choice or programming test, shown in Fig. 7. In Fig. 7 (b), the learner will receive help from the system that recommends further study after an exercise or a test, in hint form. The hint is presented in three steps to support and to remind the learner.

Next, the learner gets the total score from both types of examinations, shown in Fig. 8(a). Afterwards, if the learner clicks to get learning recommendations, they can obtain the content that they need to improve with. The instructional model in Fig. 8(b) provides content for learner from teacher (or admin). They can manage the information of learner, course and examination to record, for adapting to a learner's style with use of FCA as in section 3.2. The learner score report is provided to teacher for monitoring learning improvement. This developed system is evaluated by testing if it improves a learner's learning performance.

5. Experimental results and discussion

The experimental process is demonstrated in Fig. 9. Initially, the participants were 70 students derived from 2 groups that enrolled in Principles of Programming Course. The first learner group (35 students) is using ITS (experimental group) and the second learner group (35 students) is not using ITS (control group). This first sample group is taken from the first-year students in Information Technology Program, while the second sample group is taken from the third-year students of Industrial Management Technology at the Prince of Songkla University, Thailand. The experimental group was studying using the proposed ITS system via the developed web application, while the control group was studying in the traditional lecture approach. Next, a pretest was administered to provide baseline for improvements after learning. Afterwards, a posttest was used to evaluate improvements by comparison with pretest. In this experiment, we used three chapters of programming to be tested.

The empirical study was performed to evaluate improvements in learning performance, by testing the following hypotheses:

• 1.

  The score of posttest is better than of pretest both when using our system and when not using the system. This hypothesis would show that the learners experience positive learning performance.

• 2.

  The scores of pretest and posttest when using our system are different from those without the system. The hypothesis is that using our system improves learning performance.

To test the first hypothesis, the pretest and posttest approach is used to count the number of correct answers given to tutorial questions during tutoring and then comparison between pretest score and posttest score. This work applied paired t-tests to pretest and posttest scores, for each chapter studied, as shown in Table 3. The results show that the posttest is better than pretest, with significant difference for each chapter.

Table 3 The paired sample test results.

Group	Methods	Mean	SD	t-test	df	p-value
Using ITS	post CH1 −pre CH1	10.629	1.087	57.848	34	<.001***
	post CH2 −pre CH2	9.629	1.784	31.936	34	<.001***
	post CH3 −pre CH3	12.114	2.529	28.333	34	<.001***
Non-using ITS	post CH1 −pre CH1	5.914	2.092	16.729	34	<.001***
	post CH2 −pre CH2	4.543	1.578	17.029	34	<.001***
	post CH3 −pre CH3	4.171	1.636	15.087	34	<.001***

* p<0.05, ^** p<0.01, ^*** p<0.001.

• Open table in a new tab

To test the second hypothesis, it was necessary to compare the learning of experimental group with that of control group. A common measure of learning is learning gain. Learning gain could be measured in a number of ways, for example the number of tutorial questions a learner answers correctly or an improvement in test scores. Learning gain was measured by using a pre-test and post-test, where the same test for three chapters was taken before and after the tutoring. Test scores were then compared to establish whether there was any improvement.

$\text{learning gain}=\text{post-test score}-\text{pre-test score}$

We use the independent samples test to assess learning gain for each chapter, shown in Table 4. The results show that using our system improved learning performance from that achieved without ITS, with a significant difference for each chapter.

Table 4 The independent samples test results.

Lesson	Mean		SD		t-test	df	p-value
	Using ITS	Non-using ITS	Using ITS	Non-using ITS
CH1	10.63	5.91	1.087	2.092	11.832	51.12	<.001***
CH2	9.63	4.54	1.784	1.578	12.633	68	<.001***
CH3	12.11	4.17	2.529	1.636	15.600	68	<.001***

CH1 = chapter 1, CH2 = chapter 2, CH3 = chapter 3.

* p<0.05, ^** p<0.01, ^*** p<0.001.

• Open table in a new tab

However, the above experiment can explicitly test only the application of ITS using FCA. The proposed knowledge base with lattice structure provided a dynamic knowledge base or an amendable knowledge base supporting adaptive learning to the users. Moreover, this structure can utilize both explicit and implicit knowledge. The explicit knowledge shows in the hierarchical structure while the implicit knowledge is discovery knowledge. These advantages should be tested in future studies.

6. Conclusion

ITSs are adaptive systems that use intelligent technologies to personalize learning according to individual student characteristics. This paper applied FCA as an intelligent method based on rule-based reasoning that dynamically predicted and adapted to a learner's style to recommend topics to the learner. The FCA was used to improve the knowledge base supporting adaptive learning, which can be achieved by suitable knowledge construction. In addition, our knowledge structure provides adaptive presentation and personalized learning with the proposed adaptive algorithm to retrieve content according to individual learner characteristics. This work developed learning recommendations in ITS that dynamically predict and adapt to a learner's style by using smartphone and web applications. It implemented a C programming language tutor for general C programming teachers and learners, called Tutor C Programming, based on the web application. To demonstrate the proposed adaptive algorithm, pretest and posttest were used to evaluate suggestion accuracy of course and class for adapting to a learner's style. Moreover, this was also used with students in a real teaching/learning environment to evaluate the performance of the proposed teaching approach. The results show that the proposed system can be used to help students or learners to improve their learning.

Declarations

Author contribution statement

J. Muangprathub: Conceived and designed the analysis; Collected the data; Contributed data or analysis tools; Performed the analysis; Wrote the paper. V. Boonjing, K. Chamnongthai: Conceived and designed the experiments; Contributed data or analysis tools; Performed the analysis; Wrote the paper.

Funding statement

This work was supported by the Thailand Research Fund (TRF) and Prince of Songkla University (PSU), under grant No. TRG5880206 and SIT581125S .

Competing interest statement

The authors declare no conflict of interest.

Additional information

No additional information is available for this paper.

Acknowledgements

The authors are deeply grateful the Thailand Research Fund (TRF) and Prince of Songkla University (PSU) grant, Thailand. This research was financially supported by the TRF and PSU , under grant No. TRG5880206 and SIT581125S . The authors also gratefully acknowledge the helpful check of English language by Assoc. Prof. Dr. Seppo Karrila.

References

1. • Goyal M.
   • Yadav D.
   • Choubey A.

   E-learning: current state of art and future prospects.

   Int. J. Comput. Sci. Issues. 2012; 9 : 409-499
2. • del Puerto Paule Ruiz Ma
   • Fernández Díaz Ma Jesús
   • Ortín F.
   • Juan R.P.

   Adaptation in current e-learning systems.

   Comput. Stand. Interfaces. 2008; 30 : 62-70
3. • Klašnja-Milićević A.
   • Vesin B.
   • Ivanović Mirjana
   • Budimac Z.

   E-learning personalization based on hybrid recommendation strategy and learning style identification.

   Comput. Educ. 2011; 56 : 885-899
4. • Latham A.
   • Crockett K.
   • McLean D.

   An adaptation algorithm for an intelligent natural language tutoring system.

   Comput. Educ. 2014; 71 : 97-110
5. • Keleş A.
   • Ocak R.
   • Keleş A.
   • Gülcü A.

   Zosmat: Web-based intelligent tutoring system for teaching–learning process.

   Expert Syst. Appl. 2009; 36 : 1229-1239
6. • Kara N.
   • Sevim N.

   Adaptive learning systems: beyond teaching machines.

   Contemp. Educ. Technol. 2013; 4 : 108-120
7. • Kuo R.
   • Hsu C.K.
   • Chang M.
   • Heh J.

   A personalized webpage reconstructor based on concept lattice and association rules.

   J. Internet Technol. 2011; 12 : 1015-1024
8. • Muangprathub J.
   • Boonjing V.
   • Pattaraintakorn P.

   A new case-based classification using incremental concept lattice knowledge.

   Data Knowl. Eng. 2013; 83 : 39-53
9. • Nussbaum M.
   • Rosas R.
   • Peirano I.
   • Cardenas F.

   Development of intelligent tutoring systems using knowledge structures.

   Comput. Educ. 2001; 36 : 15-32
10. • Litman D.

    Speech and Language Processing for Adaptive Training.

    Cambridge University Press, 2013
11. • Taran T.
    • Sirota S.

    Knowledge learning technology for intelligent tutoring systems.

    Int. J. Inf. Theories Appl. 2007; 10 : 153-158
12. • Ganter B.
    • Wille R.

    Formal Concept Analysis: Mathematical Foundation.

    Springer, Heidelberg 1999
13. • Johnson W.L.
    • Valente A.

    Tactical language and culture training systems: using artificial intelligence to teach foreign languages and cultures.

    in: Proceedings of the 20th National Conference on Innovative Applications of Artificial Intelligence. 2008: 1632-1639
14. • van Seters J.R.
    • Ossevoort M.A.
    • Tramper J.
    • Goedhart M.

    The influence of student characteristics on the use of adaptive e-learning material.

    Comput. Educ. 2012; 58 : 942-952
15. • D'Mello S.K.
    • Graesser A.
    • King B.

    Toward spoken human-computer tutorial dialogues.

    Hum.-Comput. Interact. 2010; 25 : 289-323
16. • Akbulut Y.
    • Cardak C.S.

    Adaptive educational hypermedia accommodating learning styles: a content analysis of publications from 2000 to 2011.

    Comput. Educ. 2008; 58 : 835-842
17. • Hwang G.J.

    A conceptual map model for developing intelligent tutoring systems.

    Comput. Educ. 2003; 40 : 217-235
18. • Wille R.

    Formal concept analysis as mathematical theory of concepts and concept hierarchies.

    in: Lecture Notes Artificial Intelligent (LNAI-3626). 2010: 1-33
19. • Mohamed H.
    • Bensebaa T.
    • Trigano P.

    Developing adaptive intelligent tutoring system based on item response theory and metrics.

    Int. J. Adv. Sci. Technol. 2013; 43 : 1-14
20. • Priss U.

    Formal concept analysis in information science.

    Annu. Rev. Inf. Sci. Technol. 2006; 40 : 521-543
21. • Zhu G.
    • Zhang Z.

    Knowledge unit discovery for programming tutoring based on formal concept analysis.

    in: 2010 International Conference on Educational and Information Technology (ICEIT 2010). 2010: 473-479
22. • Ghadirli H.M.
    • Rastgarpour M.

    A web-based adaptive and intelligent tutor by expert systems.

    in: Advances in Computing and Information Technology. 2013: 87-95
23. • Kularbphettong K.

    An adaptive web-based intelligent tutoring using mastery learning and logistic regression techniques.

    J. Theor. Appl. Inf. Technol. 2014; 66 : 31-35
24. • Graesser A.C.
    • D'Mello S.K.
    • Hu X.
    • Cai Z.
    • Olney A.
    • Morgan B.

    Auto tutor.

    in: Applied Natural Language Processing: Identification, Investigation, and Resolution. 2012: 169-187
25. • Daomin X.

    Multi-agent based e-learning intelligent tutoring system for supporting adaptive learning.

    in: 2013 Fourth International Conference on Intelligent Systems Design and Engineering Applications. 2013: 393-397
26. • Beydoun G.

    Formal concept analysis for an e-learning semantic web.

    Expert Syst. Appl. 2009; 36 : 10952-10961
27. • Cheng S.Y.
    • Lin C.S.
    • Chen H.
    • Heh J.

    Learning and diagnosis of individual and class conceptual perspectives: an intelligent systems approach using clustering techniques.

    Comput. Educ. 2005; 44 : 257-283
28. • Chen C.

    Intelligent web-based learning system with personalized learning path guidance.

    Comput. Educ. 2008; 51 : 787-814
29. • Yang Shuqun
    • Ding S.
    • Yao Zhiqiang

    The theory about cd-cat based on fca and its application.

    Int. J. Distance Educ. Technol. 2009; 7 : 61-62
30. • Dašić P.
    • Dašić J.
    • Crvenković B.
    • Šerifi V.

    A review of intelligent tutoring systems in e-learning.

    Ann. Univ. Oradea. 2016; 2016 : 85-89
31. • Harley J.M.
    • Bouchet F.
    • Hussain M.S.
    • Azevedo R.
    • Calvo R.

    A multi-componential analysis of emotions during complex learning with an intelligent multi-agent system.

    Comput. Hum. Behav. 2015; 48 : 615-625
32. • Grawemeyer B.
    • Mavrikis M.
    • Holmes W.
    • Gutierrez-Santos S.
    • Wiedmann M.
    • Rummel N.

    Affecting off-task behaviour: how affect-aware feedback can improve student learning.

    ()2016
33. • Grivokostopoulou F.
    • Perikos I.
    • Hatzilygeroudis I.

    An educational system for learning search algorithms and automatically assessing student performance.

    Int. J. Artif. Intell. Educ. 2017; 27 : 207-240
34. • Kose U.
    • Arslan A.

    Optimization of self-learning in computer engineering courses: an intelligent software system supported by artificial neural network and vortex optimization algorithm.

    Comput. Appl. Eng. Educ. 2017; 25 : 142-156
35. • Khachatryan G.A.
    • Romashov A.V.
    • Khachatryan A.R.
    • Gaudino S.J.
    • Khachatryan J.M.
    • Guarian K.R.
    • Yufa N.V.

    Reasoning mind genie 2: an intelligent tutoring system as a vehicle for international transfer of instructional methods in mathematics.

    Int. J. Artif. Intell. Educ. 2014; 24 : 333-382
36. • Weragama D.
    • Reye J.

    Analysing student programs in the php intelligent tutoring system.

    Int. J. Art. Intell. Educ. 2014; 24 : 162-188
37. • Soh L.K.
    • Khandaker N.
    • Jiang H.

    I-minds: a multiagent system for intelligent computer-supported collaborative learning and classroom management.

    Int. J. Art. Intell. Educ. 2008; 18 : 119-151
38. • Latham A.
    • Crockett K.
    • McLean D.
    • Edmonds B.

    A conversational intelligent tutoring system to automatically predict learning styles.

    Comput. Educ. 2012; 59 : 95-109
39. • Wang D.
    • Han H.
    • Zhan Z.
    • Xu J.
    • Liu Q.
    • Ren G.

    A problem solving oriented intelligent tutoring system to improve students' acquisition of basic computer skills.

    Comput. Educ. 2015; 81 : 102-112
40. • Ganter B.
    • Wille R.

    Applied Lattice Theory: Formal Concept Analysis.

    Institute for Algebra, 1997
41. • Dabolinš J.

    Trends of the usage of adaptive learning in intelligent tutoring systems.

    in: Databases and Information Systems (Baltic DB&IS 2012): the 10th International Baltic Conference. 2012: 191-196
42. • Desmarais M.C.
    • Baker R.S.J.

    A review of recent advances in learner and skill modeling in intelligent learning environments.

    User Model. User-Adapt. Interact. 2012; 22 : 9-38
43. • Phobun P.
    • Vicheanpanya J.

    Adaptive intelligent tutoring systems for e-learning systems.

    Proc., Soc. Behav. Sci. 2010; 2 : 4064-4069
44. • Kumar C.
    • Singh P.

    Knowledge representation using formal concept analysis: a study on concept generation.

    in: Global Trends in Intelligent Computing Research and Development. 2014: 306-336
45. • Singh P.K.
    • Kumar C.
    • Gani A.

    A comprehensive survey on formal concept analysis, its research trends and applications.

    Int. J. Appl. Math. Comput. Sci. 2016; 26 : 495-516

Article Info

Publication History

Published online: October 23, 2020

Accepted: October 8, 2020

Received in revised form: July 3, 2020

Received: April 5, 2020

Identification

DOI: https://doi.org/10.1016/j.heliyon.2020.e05227

Copyright

© 2020 The Author(s). Published by Elsevier Ltd.

User License

Creative Commons Attribution – NonCommercial – NoDerivs (CC BY-NC-ND 4.0) |

How you can reuse

Permitted

For non-commercial purposes:

• Read, print & download
• Redistribute or republish the final article
• Text & data mine
• Translate the article (private use only, not for distribution)
• Reuse portions or extracts from the article in other works

Not Permitted

• Sell or re-use for commercial purposes
• Distribute translations or adaptations of the article

Elsevier's open access license policy

ScienceDirect

Access this article on ScienceDirect

• View Large Image
• Download Hi-res image
• Download .PPT

Related Articles

Cell Press commenting guidelines

To submit a comment for a journal article, please use the space above and note the following:

• We will review submitted comments within 2 business days.
• This forum is intended for constructive dialog. Comments that are commercial or promotional in nature, pertain to specific medical cases, are not relevant to the article for which they have been submitted, or are otherwise inappropriate will not be posted.
• We recommend that commenters identify themselves with full names and affiliations.
• Comments must be in compliance with our Terms & Conditions.
• Comments will not be peer-reviewed.

Design Recommendations for Intelligent Tutoring Systems Volume 1

Source: https://www.cell.com/heliyon/fulltext/S2405-8440(20)32070-3

Share this post