A Parallel Education Based Intelligent Tutoring

Systems Framework
Sifeng Jinga,b, Ying Tanga,b,c, Senior Member,IEEE, Xiwei Liua,b,*, Xiaoyan Gonga,b,Wei Cuid Joleen Liangd,Member,IEEE

a. The State Key Laboratory for Management and Control of Complex Systems,
Institute of Automation, Chinese Academy of Sciences
Beijing 100190, China
b. Institute of Smart Education Systems, Qingdao Academy of Intelligent Industries
Qingdao 266044, China
c. Department of Electrical and Computer Engineering, Rowan University,
Glassboro, NJ 08028, USA
d. Squirrel AI Learning
Shanghai 200030, China
[email protected]

Abstract—although online education has improved efficiency 15% higher than that of students who did not use ALEKS [4].
of learners' access to high-quality educational resources, real- A similar result was found for ALEKS chemistry [5]. A recent
time interaction between instructors and learners have not yet study compared the performance of the ITS developed by
been achieved in online personalized learning. Intelligent Squirrel AI with that of human teachers and found the former
Tutoring Systems (ITS) provides a feasible way to realize real- outperformed the latter [6-9].
time personalized learning guidance and resource
recommendations by applying AI to capture and analyze online Key technologies of developing ITS consist of domain
learners’ characteristics and behaviors. In this paper, reviews knowledge graph, learner model, guidance mechanism and
and trends of ITS are discussed and three challenges of ITS human-computer interaction mechanism. Comprehensive
research &development are pointed out: learner model, guidance learner model, precise decision-making and guiding, and
mechanism and human-computer interaction mechanism. In personalized interactions between human and computers are
order to address these issues, parallel intelligence theory is the key projects to promote performance of ITS.
introduced and a parallel intelligence education based ITS
framework is proposed. In terms of learner model, although there are many types
of existing learner models, they are all models modeling one or
Keywords—Intelligent Tutoring System, Personalized education, two learner’s characteristics. Some models predict learner’s
Online Education, Metacognition, Knowledge Graph mastery on domain knowledge, some models predict learner’s
learning styles, and some models predict learner’s emotions.
I. INTRODUCTION These models were built by various types of online data, such
E-learning has been massively adopted worldwide and it as data of learning assessment reports, learning behaviors and
provides learners efficient access to high-quality educational learning habits [10, 11]. And there are few models that take
resources. Learners can study in their own time, at their own into account learners’ comprehensive characteristics, such as
pace and in their preferred space in e-learning environment. knowledge, learning style, cognitive style and social
However, given the lacks of real-time interactions between communication behaviors.
instructors and learners in e-learning, some learners often feel Furthermore, few models have been built on integration of
lonely and confused during learning process in e-learning learner’s knowledge, cognitive domain and affective domain.
environments. How to achieve effective and large-scale one-to- Learner models based on cognitive style can describe internal
one coaching has become focus of attention in e-learning. cognition of learners, while models based on learning style and
Intelligent Tutoring System (ITS) proposes a feasible way to emotions characterize externalities of learning behaviors.
deal with those issues by applying AI technologies to identify Establishing learner model that integrates cognitive style,
learning behaviors and thus provide personalized support for learning style and emotions can describe learners’
learners. Hence, ITS has attracted increased attention of comprehensive characteristics.
researchers.
In terms of guidance mechanism, knowledge-based
ITS was formally proposed by Sleepman and Brown in recommendation methods have been adopted to implement
1982, aiming to imitate human teachers and provide learners personalized knowledge recommendation, which has promoted
with “just-in-time” and individualized instruction in e-learning performance of personalized guidance. Little efforts have been
environments by adapting to learners’ knowledge, preferences, done on realization of learners’ personalized guidance based on
and other needs [1-3].Many ITS have shown their effectiveness learner's cognitive characteristics, knowledge, and knowledge
for assisting student learning. For example, the research found graph [12, 13]. Cognitive laws and knowledge are both
that the curriculum pass rate of students who used ALEKS important factors that jointly describe learning processes.
mathematics, one of the popular ITS in the United States, was
*Corresponding author

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

Authorized licensed use limited to: Auckland University of Technology. Downloaded on December 22,2020 at 22:33:55 UTC from IEEE Xplore. Restrictions apply.
Therefore, it is of great practical significance to establish Therefore, e-learner's emotion model has received more and
comprehensive learner model and study guidance mechanism more attention. According to time granularity, existing emotion
based on learners’ cognitive style and knowledge. models can be divided into snapshot emotion models and
continuous emotion models. According to the way of model
In terms of interaction mechanisms, existing interaction data collection, existing emotion models can be divided into
mechanism mostly involve direct support (sometimes called subjective emotion models and objective emotion models.
“homework helpers”), which means that hints are provided
upon learners’ requests [3, 14, 15]. But all of them rarely Learner comprehensive model refers to a holistic model
considered comprehensive factors such as metacognition, built by considering multiple aspects of a learner's
emotion and collaboration. ITS is not only a tool to help characteristics [28, 29]. For example, in adaptive learning
learners acquire knowledge and facts, but also a teacher to help system, ability of adaptive learning system to adapt to learners
learners perform metacognition tasks. Metacognition is can be improved by comprehensively modeling learners'
recognition and thinking of cognitive activities. By using ITS knowledge level, cognitive ability, and preferences information
to monitor and analyze learners' learning strategies, problem [27, 30]. Recently, learner modeling techniques has evolved
solving strategies, and collaborative learning strategies, with application of artificial intelligence technology in field of
learners can understand themselves more at methodological smart education. Semi-automatic or fully automatic data-driven
level [16]. Therefore, constructing interactive instruction modeling methods have gradually replaced traditional manual
libraries that comprehensively consider cognition / modeling methods. Deep knowledge tracing model is a typical
metacognition, emotion, and collaboration is of great value in case. Thereby, constructing learner models will rely more on
improving performance of personalized guidance systems. intelligent programs and large amounts of data, rather than
expert domain knowledge [31-33]. In recent years, deep
So, in next chapter, this paper will go over ITS in more
details and introduce challenges and issues. Chapter 3 learning-based artificial intelligence technologies have greatly
summarize progress and trends, in chapter 4, parallel improved utilization of large-scale data and learning ability of
intelligence is introduced and parallel education-based, brain intelligent models, which will provide technology support for
inspired and data-driven personalized ITS framework is building more powerful learner models [34].
proposed to address those issues. Chapter 5 comes to 2.2 Guidance mechanism
conclusion and perspectives. How to arrange learning content, exercises, and activities
II. RELATED WORKS according to each learner’s characteristics to achieve the
learning objectives in e-learning environment is the issue
In this section, learner model, guidance mechanism and discussed in this part.
human-machine interaction mechanism of ITSs were reviewed
respectively. To address this issue, constructing domain knowledge
graph and developing appropriate recommendation strategies
2.1 Learner model based on comprehensive learner models are two main
Learner model stores static and dynamic data of learners. challenges.
ITS uses these information to predict learners’ learning style Knowledge is important for human existence, while data is
and behavior, and thereby provide learners with tailored incomprehensible. Knowledge representation is an important
feedback and personalized guidance [17, 18]. A learner model technology that can discover correlations across different
incorporates learners’ current knowledge and personal traits to objects from different data sources, construct semantic
construct a learner profile. Learning style, cognitive style, connections between these objects, and finally obtain usable
working memory capacity and personality type are four knowledge. So knowledge graph came into being [35]. Big
personal traits that have attracted the interest of researchers in data of different types can be turned into valuable information
recent years. Learner model is a basis of learning analysis and that both humans and machines can understand based on
data mining [19]. Currently, established learner models mainly knowledge graph. Learning and reasoning are two core
include learner knowledge models, cognitive models, technologies of knowledge graph [36].
emotional models, and learning behavior models [20]. Analysis
of these models showed that learner behavior, cognitive style, Intelligent learning technology allows knowledge graphs-
and emotion were three important factors that characterize based domain knowledge database to be updated and improved
learners, and that there were interactions between learner timely. Intelligent reasoning technology makes knowledge
behavior, cognition, and emotion [21, 22]. graph-based knowledge bases more convenient for people to
query, retrieve and call. At the same time, people often
Learner's assessment data and learning trajectory data were combine natural language processing and machine learning to
often used to construct learner's knowledge state model. construct knowledge graph in different domains, which has
Generally, appropriate features can be selected to group greatly improved efficiency and effect of knowledge graph
learners by learner's behavioral data. Some studies tried to construction [37-40]. Thereby knowledge graph was applied to
quantify learners' cognitive behavior from a certain perspective construction of artificial intelligence (AI) subject knowledge
based on behavioral data [23, 24]. base of our ITS based on our previous works [41-43].
Emotions play a key role in learning process [25]. ITS can guide students to achieve learning objects of
According to constructivist theory, positive emotions help corresponding subjects according to learner’s characteristics.
learners successfully achieve their learning goals [26, 27]. Personalized instruction is implemented by being able to adapt

Authorized licensed use limited to: Auckland University of Technology. Downloaded on December 22,2020 at 22:33:55 UTC from IEEE Xplore. Restrictions apply.
to the knowledge, learning abilities and needs of each research from Jean Plaget and Lev Vygotsky, learning is not
individual learner. And this function is implemented by only derived from learner's own exploration and acquisition of
developing appropriate recommendation strategies. The knowledge, but also it has a close relationship with external
recommendation system was proposed in the context of big social and cultural environment. If learners have a deep self-
data. It was initially successfully applied to online e- knowledge in learning process, their learning strategies and
commerce, information retrieval, mobile applications, and life self-control in the learning process will be greatly improved.
services. Recommendation system has received widespread This cognitive process is what we call metacognition.
attention and research to solve information overload in Metacognition refers to conscious aware and control of
academic and industrial circles [44]. Recommendation learner’s learning. Metacognitive approaches enable learners to
algorithm is the core of recommendation system, which helps monitor their progress when they engage in understanding and
users easily find objects from massive data by mining binary learning something. Generally, learners can improve their
relationship between users and items, and generates learning efficiency and performance by improving their
personalized recommendation lists to meet user preferences. learning strategies [64-66].The metacognitive skill puts
There are five main types of recommendation algorithms [45]: forward the centrality of subjective judgment of confidence
content-based recommendations, knowledge-based and coping strategies in the success of performing cognitive
recommendations, collaborative filtering recommendation tasks [67].Although more and more researchers have been
algorithms, and hybrid recommendation algorithms. aware that metacognitive ability is of great significance to the
Collaborative filtering approach has been proven to have the improvement of learners’ learning proficiency, metacognitive
potential to improve performance of a learning path models is rarely involved in ITSs now.
recommendation in ITS [46]. Collaborative filtering (CF)
techniques, content-based filtering (CBF) techniques and III. PROGRESS AND TREND ANALYSIS OF ITS
context-aware filtering (CAF) techniques are three common From research &development trends of ITSs described
collaborative filtering approaches. Generally, the user-based above, it can be seen that three major issues need to be
CF algorithm recommend items by feedback of users with resolved to improve their performance. Firstly, most existing
similar preferences to the target user, while the CBF algorithm learner models are based on learning data and can predict one
operate with the similarity of the items, so similar items to the of learner’s characteristics. And few learner models take into
ones liked by the target user are recommended. The location- cognition, affective, behavior or psychomotor and mixed into
based CAF methods use the location of users to recommend account. Secondly, ITS's guidance mechanism focused on
items close to them. The challenges of these techniques are guidance for learning strategies, learning paths, and knowledge.
presented namely: how to find similar users in order to And less consideration is given to learner’s cognitive
consider their feedback when generating new characteristics and metacognitive ability. Few researches have
recommendations in the CF techniques, how to create user’ conducted on guidance mechanisms that integrate learner’
profiles and classify items in the CBF techniques and how to cognitive characteristics，emotion, behavior and knowledge
use the location and the tracking of users in the CAF states. Thirdly, ITS's human-machine interaction mechanism
approaches [47]. focused on various types of knowledge display.
Overall, most Chinese scholars have implemented Comprehensive factors of cognitive \ metacognitive and
personalized knowledge recommendation based on knowledge- emotional \ social have not been considered in interactive
based recommendation strategies [48,14,15,49], which has instruction libraries of existing ITSs.
promoted performance of personalized guidance systems. To improve existing ITSs’ performance, A framework for
However, there is little research on how to combine students' ITS is proposed based on parallel education theory. Firstly, a
cognitive style with learning objectives of the corresponding comprehensive learner model will be explored and constructed
disciplines to achieve personalized guidance for students. To based on multi-mode learning data. Secondly, precise guidance
this end, a knowledge graph-based personalized guidance mechanism will be studied based on domain knowledge graph
mechanism that combines students' cognitive characteristics and recommendation algorithms. Finally, basic metacognitive
with domain knowledge is mainly explored in on-going work. skills of learners needs to be took into account in human-
2.3 Human-machine interaction mechanism machine interaction mechanisms.
Currently, most adaptive ITSs interact with learners based IV. THE MAIN FRAMEWORK OF ITS BASED ON PARALLEL
on learners’ knowledge needs [50]. In many ITSs, a series of EDUCATION
pre-tests have been designed. Learners' answers to these tests
confirmed their knowledge level [51-54]. In some ITSs, natural 4.1 Parallel Education Theory
language processing techniques were used to imitate learner's With the radical and transformative technological
dialogue with virtual tutors, thereby learner's knowledge needs advancement, parallel intelligent education came into being. It
were reflected [55-59]. empowers the next generation of digital citizens in more
To improve performance of ITSs, more and more effective and efficient learning. Intelligent tutoring system is
researchers have begun to take learners’ emotion into account one of these tools [68]. To improve existing ITSs’
when modeling learner learning behavior in the last ten years performance, A framework for ITS is proposed based on
[60]. Many ITSs have provided personalized learning guidance parallel education theory.
adapt to learners’ emotion changes [61-63]. According to

Authorized licensed use limited to: Auckland University of Technology. Downloaded on December 22,2020 at 22:33:55 UTC from IEEE Xplore. Restrictions apply.
 According to parallel education theory [69, 70], parallel Research contents mainly are listed as following. How
implementation of artificial education system and real pedagogy is applied to education and learning process.
education system is used to improve personalized guidance Learners’metacognition and cognitive law. How to use brain
function of intelligent tutoring system. The overall idea of the science to conduct sentiment analysis. How to use computer
science to achieve personalized guidance, tailored feedback
project research is shown in Figure 1. Based on real teaching
and human-computer interaction of ITS. Rules of learner
system, a artificial teaching system was constructed to model cognition and learning. Relationship between learner and tutor.
learning,evaluating and guiding process of students learning in Relationship between learner and learner. Relationship
online learning environment. The real teaching system between learner and ITS environments. Relationship between
provides data and test for the artifical teaching system,while man and machine.
the artificial teaching system provides theory guidance for
As such, research plan has the following three objectives:
improving ITS’perfomance.
1) Key technologies of ITS will be innovated,
4.2 Research Plan comprehensive learner models will be built and precise
Based on general idea of project research shown in Figure guidance mechanisms will be formed based on parallel
1, a research plan to improve personalized guidance function education theories. Firstly, learner models will be built and
of intelligent tutoring system is shown in Figure 2. optimized based on cognition law, metacognition ability and
knowledge level of learners in e-learning environments.
Secondly, a precise guidance mechanism will be built based on
knowledge graph and recommendation algorithms based on
comprehensive learner models. Thirdly, a personalized human-
computer interaction instruction library and virtual learning
partners will be built through interaction mechanisms
integrating cognitive / metacognitive with emotional / social
factors. Finally, Multi-modal personalized human-computer
interaction instruction scenarios will be implemented.
2) Key technologies of ITS will be deployed and
implemented on basis of comprehensive learner models,
precise guidance mechanism and interaction mechanisms
described above. A comprehensive learner model agent and a
precise guidance mechanism agent will be developed
respectively. Combined with the K12 artificial intelligence
Figure.1 General idea of project research curriculum system[71,72] developed by Institute of Smart
Education Systems, Qingdao Academy of Intelligent Industries,
a domain knowledge graph will be built, and then a
personalized human-computer interaction instruction library
and an adaptive interaction engine agent will be developed
based on existing VR-based learning system [73]. Agents and
knowledge graph described above will be incorporated to build
a VR-based, personalized K12 artificial intelligence course
demonstration system.
3) After demonstration system is set up, system
demonstration application and model algorithm optimization
will be performed. Data of learners’ characteristics will be
collected and analyzed in different regions and different
cultural groups from pilot schools in the United States and
China. And then, system testing and model optimization will
be conducted. Application and optimization of personalized
Figure.2 Research Plan guidance demonstration system for artificial intelligence
education courses for K-12 will be carried out, which will lay a
ITS is an integrated system that integrates theoretical
foundation for future promotion of ITS.
research (cognitive science, learning science), technological
development (computer science, computer engineering), V. CONCLUSIONS AND AND PERSPECTIVES
educational practice (teaching methods, teaching strategies),
and empirical evaluation. Therefore, the proposed research ITS can provide tailored feedback and personalized
plan combines pedagogy, psychology, brain science, computer guidance for online learners by capturing learner’s individual
science, computer engineering and cyber-physical-social- characteristic. Therefore, it has been paid more and more
education to address challenges and issues in ITS research attentions by many researchers in smart education filed in
&development. recent years. Development trend of ITS was fully summarized
and analyzed in this article. Results revealed that learner

Authorized licensed use limited to: Auckland University of Technology. Downloaded on December 22,2020 at 22:33:55 UTC from IEEE Xplore. Restrictions apply.
models, precise guidance mechanism and human-machine based on Knowledge Hierarchical Graph,ReKHG[J]. Computer
interaction mechanism were three major issues that ITS faced. Engineering and Applications, 2018, 54(10): 229-235.
In order to address these issues, parallel intelligence theory was [16] C. L. James, Y. H. Eun, Y. L. Seung, et al., Serious Games Get Smart :
Intelligent Game-Based Learning Environments. Aimagazine, Winter
introduced and a parallel education based ITS framework was 2013:31-45.
proposed. And then a systematic research plan was described. [17] ZHUANG Kejun,HE Baoxun. The Modeling Research in Personalized
The innovation of this solution were to build a comprehensive Recommendation System Based in Network Learning[J]. China
learners model based on cognition law/ metacognition law, Educational Technology & Equipment, 2015(18): 67-69.
knowledge level and emotions, to explore precise guidance [18] Ding Jihong,Liu Huazhong. Construction of a Learner Model Based on
mechanism based on knowledge graph and to build a multi- the Influence Factors of Educational Resources ' Choice[J]. Distance
mode human-machine interaction mechanism based on Education Journal, 2017. 35(04): 97-103.
cognition \ metacognition, emotion\social. [19] Zhang Ting.Research on the Personalized Learner Model of Modern
Distance Education[M].2017,JIANGNAN UNIVERSITY.
The focal points of ongoing research are to identify the [20] Sun Li,Zhang Ting. Design and Analysis of Personalized Learner Model
personal traits used for learner model and to compare the under the Online Diploma Education[J], Distance Education Journal,
performance of key identification techniques used for learner 2017. 35(03): 93-101.
modeling. [21] Zhu Bin, Wen zhikang.Research and Application of Learner Model
Standard. Computer Engineering,2002. 28(z1): 300-303.
REFERENCES [22] Chang T Y , Ke Y R . A personalized e-course composition based on a
genetic algorithm with forcing legality in an adaptive learning system[J].
Journal of network and computer applications, 2013, 36(1):533-542.
[1] Chen, P.H., et al., KnowEdu: A System to Construct Knowledge Graph
[23] WANG Jue,XIE Yueguang. Research on the Construction of Precision
for Education. IEEE Access, 2018. 6: p. 31553-31563.
Learner Model in the Intelligent Learning Environment[J].Open
[2] D. H. Sleeman, J. S. Brown, Intelligent Tutoring Systems, New Education Research, 2017. 23(06): 104-110+59.
York: Academic Press, 1982: 1-11.
[24] JM. Ye, S. Xu., DX..Luo, et al, Research on the Construction and
[3] H. C. Lane, Intelligent tutoring systems: Prospects for guided practice Application of Individual Learner Model. Procedia Computer Science,
and efficient learning, Whitepaper for the Army's Science of Learning 2018. 131.
Workshop, Hampton, VA, 2006.
[25] B. Jla., YJ.Yang, J. Zhang, Study on Learner Modeling in Adaptive
[4] S. Mojarad, A. Essa, S. Mojarad, & R. S. Baker, 2018. Studying Learning System. Journal of Computers, 2012. 7(10).
adaptive learning efficacy using propensity score matching. In
[26] ZHANG Xiao-qing,XIAO Jian-hua. Algorithms Design of Intelligent
Companion proceedings 8th international conference on learning
Tutoring System Based on Student Cognizance Model[J]. Computer
analytics & knowledge (pp. 1–8). Sydney: ACM.
Knowledge and Technology, 2008. 4(30): 216-217+223.
[5] M. Richards-Babb, R. Curtis, B. Ratcliff, A. Roy, T. Mikalik, General
[27] I. Gakhal, S. Bull, An open learner model for trainee pilots. ALT-J,
Chemistry Student Attitudes and Success with Use of Online
2008. 16(2).
Homework: Traditional-Responsive versus Adaptive-Responsive. J.
Chem. Educ. 2018, 95 (5), 691 [28] XU Zhenguo,ZHANG Guanwen,MENG Xiangzeng, et al. Learners’
Emotion Recognition and Its Application Based on Deep Learning[J]. E-
[6] H. Li, W. Cui, Z. Xu, Z. Zhu and M. Feng, 2018. Yixue adaptive
education Research, 2019. 40(02): 87-94.
learning system and its promise on improving student learning.
Proceedings of the 10th International Conference on Computer [29] Xie Zhongxin,Wang Linquan,Ge Yuan. Construction of the Congnitive
Supported Education. Student Model in an ICAI[J]. Computer Engineering and Applications,
2005. 41(3): 229-232.
[7] M. Feng, W. Cui, and S. Wang, 2018, Adaptive learning goes to China.
Artificial Intelligence in Education, 10948, 89-93. [30] YE Junmin,HUANG Pengwei,WANG Zhifeng,et al. Research on
Learning Effect Evaluation of Model Building Based on Individual
[8] W. Cui, Z. Xue, K.-P. Thai, “Performance comparison of an AI-based
Learner[J]. E-education Research, 2018. 39(10): 90-96.
Adaptive Learning System in China.” Chinese Automation Congress,
2018, Xi’an: IEEE [31] Feng Hong,Wang Lu,Yang Hui.The Research of Multi-purposes Based
Learner Model[J]. Computer Engineering and Applications, 2003.
[9] Wei Cui, Zhen Xue, Jun Shen, Geng Sun, Jianxin Li, “The Item 39(36): 98-100.
Response Theory Model for an AI-based Adaptive Learning System.”
18th International Conference on Information Technology Based Higher [32] Ding Rongtao. To Construct the Disaggregated Model of the Higher
Education and Training, 2019, Magdeburg: IEEE Vocational Student's Network Learning Based on Decision Tree[J].
Journal of Distance Education, 2010. 28(5): 22-26.
[10] Anohina, Advances in Intelligent Tutoring Systems: Problem-solving
Modes and Model of Hints, International Journal of Computers, [33] LIU Danqing,LI Yazhi,WU Zhenqiang, et al. The optimal control
Communication and Control, 2007, 2(1):48-55. strategy in continuous teaching based on the compound memory-
forgetting model frame[J]. Journal of Shaanxi Normal
[11] S. A. Hosseint, H. Jahankhani, M. Yarandi, et al., Towards an
University(Natural Science Edition), 2018. 46(4): 18-26.
Ontological Learners’ Modelling Approach for Personalised e-Learning.
International Journal of Emerging Technologies in Learning (iJET), [34] LIU Haijiao. Research on Evaluation Model of Network Teaching
2013, 8(2):100-120. Based on Multi-Agent[J]. Computer Programming Skills &
Maintenance, 2012(2): 112-113.
[12] B. Jla, Y. Yang, J. Zhang, Study on Learner Modeling in Adaptive
Learning System. Journal of Computers, 2012 , vol. 7 ,No. 10, pp:100- [35] Sheth, Semantics empowered Web 3.0:managing enterprise, social,
120. sensor, and cloud-based data and service for advanced applications.
San Rafael, CA: Morgan and Claypool, 2013.
[13] S. Shishehchi, S. Y. Banihashem, et al., A Proposed Semantic
Recommendation System for E-Learning: A Rule and Ontology Based [36] WU Yun-bing,YANG Fan,LAI Guo-hua, et al. Research Progress of
E-Learning Recommendation System. Proc. of Int’l Sym. in Information Knowledge Graph Learning and Reasoning[J]. Journal of Chinese
Technology. IEEE CS (1), 2014:1-5.. Computer Systems, 2016,37(9): 2007-2013.
[14] Xie Zhenping,Jin Chen,Liu Yuan, Personalized Knowledge [37] HUANG Huasheng. Study on Personalized Learning Resource
Recommendation Model Based on Constructivist Learning Theory[J]. Recommendation Based on Knowledge Graph[J]. Software
Journal of Computer Research and Development,2018, 55(1): 125-138. Engineering,2018,21(10)：24-27.
[15] JIANG Changmeng,FENG Jun,SUN Xia,CHEN Jing,ZHANG [38] Shah P , Mayer R E , Hegarty M . Graphs as aids to knowledge
Lei,FENG Hongwei. Personalized exercises Recommendation algorithm construction: Signaling techniques for guiding the process of graph

Authorized licensed use limited to: Auckland University of Technology. Downloaded on December 22,2020 at 22:33:55 UTC from IEEE Xplore. Restrictions apply.
 comprehension.[J]. Journal of Educational Psychology, 1999, 91(4):690- [57] G. Alobaidi., K. A. Crockett, J. D. O’Shea., et al, Abdullah: An
702. intelligent Arabic conversational tutoring system for modern Islamic
[39] Wang Q , Mao Z , Wang B , et al. Knowledge Graph Embedding: A education. Paper presented at the Lecture Notes in Engineering and
Survey of Approaches and Applications[J]. IEEE Transactions on Computer Science.
Knowledge and Data Engineering, 2017, 29(12):2724-2743. [58] W. Kacalak, M. Majewski, J. M. Zurada, Intelligent E-learning systems
[40] Jin Guiyang,Lü Fuzai,Xiang Zhanqin. Enterprise information integration for evaluation of User’s knowledge and skills with Efficient information
based on knowledge graph and semantic web technology[J]. Journal of processing. In L. In, R. Rutkowski, R. Scherer, L. A. Tadeusiewicz, J.
Southeast University(Natural Science Editi, 2014，44(2) : 250-255． M. Zadeh, Zurada (Eds.), Artificial intelligence and Soft Computing
(Vol. 6114, pp. 508). Heidelberg, Berlin: Springer.
[41] Wang F Y . Software-defined systems and knowledge automation: a
parallel paradigm shift from newton to merton[J]. Zidonghua [59] Graesser A C , Chipman P , Haynes B C , et al. AutoTutor: an intelligent
Xuebao/Acta Automatica Sinica, 2015, 41(1):1-8. tutoring system with mixed-initiative dialogue[J]. IEEE Transactions on
Education, 2005, 48(4):612-618.
[42] ZENG Shuai,WANG Shuai,YUAN Yong,et,al. Towards Knowledge
Automation: A Survey on Question Answering Systems[J]. Acta [60] Normadhi, Nur & Shuib, Liyana & Nasir, Hairul & Bimba, Andrew &
Automatica Sinica,2017，43(9): 1491-1508. Idris, Norisma & Balakrishnan, Vimala. (2018). Identification of
personal traits in adaptive learning environment: Systematic literature
[43] LIU Shuo,WANG Shuai,FU Huanzhang,WANG Feiyue. Software- review [J]. Computers & Education,2019,130:168-190.
Defined Crime Scene Analysis Process and Its Knowledge Automation
Scheme[J].Pattern Recognition and Artificial Intelligence, 2016, 29(10): [61] J. M. Harley, F. Bouchet, M. S. Hussain, et al, A multi-componential
876-883. analysis of emotions during complex learning with an intelligent multi-
agent system. Computers in Human Behavior, 48, 615–625.
[44] HUANG Zhen-Hua,ZHANG Jia-Wen,TIAN Chun-Qi,et,al. Survey on doi:10.1016/j.chb.2015.02.013.
Learning-to-Rank Based Recommendation Algorithms[J]. Journal of
Software, 2016,27(3): 691-713. [62] B. Grawemeyer, M. Mavrikis, W. Holmes, et al., Affecting Off-task
Behaviour: How affect-aware feedback can improve student learning.
[45] Zhou chunhua, Shen jianjing, Li yan,et,al.Review of Classical Paper presented at the ACM international Conference Proceeding Series.
Recommendation Algorithms[J]. Computer Science and Application,
2019,9(09):1803-1813. DOI:10.12677/CSA.2019.99202. [63] Costello, Robert, Adaptive intelligent personalised learning (aipl)
environment (U621351 Ph.D.), University of Hull (United Kingdom),
[46] Milicevic A K , Vesin B , Ivanovic M , et al. E-Learning personalization Ann Arbor. Retrieved from https://search.proquest.com/docview
based on hybrid recommendation strategy and learning style /1654740829?accountid=41304ProQuest Dissertations & Theses A&I;
identification[J]. Computers & Education, 2011, 56(3):885-899. ProQuest Dissertations & Theses Global database.
[47] Design of a recommender system based on users’ behavior and [64] M. Samarakou, P. Prentakis, D.Mitsoudis, et al, Application of fuzzy
collaborative location and tracking[J]. Journal of Computational Science, logic for the assessment of engineering students. Paper presented at the
2016, 12:83-94. IEEE Global engineering education Conference, EDUCON.
[48] Zhao Wei,Jiang Qiang,Wang Pengjiao,et,al. Ontology-driven
[65] K. Chrysafiadi, M. Virvou, Evaluating the integration of fuzzy logic into
Personalized Recommendation of Knowledge Resource in e-Learning[J].
the student model of a web-based learning environment. Expert Systems
China Educational Technology, 2015, (05): 84-89.
with Applications, 39(18), 13127–13134.
[49] ZHANG Bingqi. A Collaborative Filtering Recommendation Algorithm doi:10.1016/j.eswa.2012.05.089.
Based on Domain Knowledge[J]. Computer Engineering,
[66] Z. Jeremic, J. Jovanovic, et al, Evaluating an intelligent tutoring system
2005,31(21):7-9,33. DOI:10.3969/j.issn.1000-3428.2005.21.003.
for design patterns: The DEPTHS experience. Journal of Educational
[50] E. Mousavinasab, N. Zarifsanaiey, N. Kalhori, et al, Intelligent tutoring Technology & Society, 12(2), 111.
systems: a systematic review of characteristics, applications, and [67] Tsai, Ya-hsun, Lin, et,al. The effects of metacognition on online
evaluation methods, Interactive Learning Environments, DOI: learning interest and continuance to learn with MOOCs[J]. Computers &
10.1080/10494820.2018.1558257. Education, 2018.
[51] S. J. Hsieh, Y. Cheng, Algorithm and intelligent tutoring system design [68] Ying Tang, Joleen Liang, Ryan Hare, and Fei-Yue Wang. A
for programmable controller programming. International Journal of Personalized Learning System for Parallel Intelligent Education[J].
Advanced Manufacturing Technology, 71(5–8), 1099–1115. IEEE Trans. on Computational Social Systems, 2020, 7(2): pp. 352-361.
doi:10.1007/s00170-013-5539-z.
[69] Wang, Fei-Yue,Tang Ying, Liu Xiwei, Yuan Yong. Social Education:
[52] K. Dolenc, B. Aberšek, intelligent and adaptive e-learning system: Opportunities and Challenges in Cyber-Physical-Social Space. IEEE
Integration into technology and science classrooms in lower secondary Transactions on Computational Social Systems[J].2019,6(2) :p 191-196,
schools. Computers and Education, 82, 354–365. April 2019.
doi:10.1016/j.compedu. 2014.12.010
[70] Wang, Fei-Yue. How to cultivate artificial intelligence talents: From
[53] F. Grivokostopoulou, I. Perikos , I. Hatzilygeroudis, An educational parallel teaching to smart education[J].Science &Technology Review,
system for learning search algorithms and Automatically Assessing 2018，36（11）：9-12.
student performance. International Journal of Artificial Intelligence in
Education, 27(1), 207–240. doi:10.1007/s40593-016-0116-x. [71] Liu, Xiwei, Dong, Xisong, Wang, Fei-Yue, Sun,Rui, Y. Gao, Y. Zhang,
J. Zhou, and X. Deng. A new framework of science and technology
[54] J. McDonald, A. Knott, S. Stein, et al, An empirically-based, Tutorial
innovation education for k-12 in Qingdao, China. 2017 ASEE
dialogue system: Design, implementation and evaluation in a first year International Forum, June 28, 2017. 2017. Columbus, OH, United states:
health sciences course. Paper presented at the 30th Annual conference
on Australian Society for computers in learning in Tertiary education, [72] Gong Xiaoyan, Zhao Liang, Tang RenHao, Guo YaNa, Liu, Xiwei, He
ASCILITE 2013. Jian, Wang Fei-Yue, Tang Ying et.al. AI Educational System for
Primary and Secondary Schools, China. 2019 ASEE International
[55] M. Dzikovska, N. Steinhauser, E. Farrow, et al, BEETLE II: Deep
Forum, June 16-19, 2019. 2019. Tampa Convention Center, Tampa,
natural language understanding and automatic feedback generation for Florida: American Society for Engineering Education.
intelligent tutoring in basic electricity and electronics. International
Journal of Artificial Intelligence in Education, 24(3), 284–332. [73] Ying Tang, Christopher Franzwa, Talbot Bielefeldt, Kauser Jahan,
doi:10.1007/s40593-014-0017-9. Marzieh Saeedi-Hosseiny, Nathan Lamb, and Shengtao Sun, “Sustain
City – Effective Serious Game Design in Promoting Science and
[56] JM. Liu., Y. Xiang., MH. Yang, Construction and application of new Engineering Education,” in Design, Motivation, and Frameworks in
intelligent MOOC teaching system based on deep learning neural Game-Based Learning, IGI Global, 2017
network in remote sensing course. Paper presented at the Proceedings –
2016 8th International Conference on Information Technology in
Medicine and Education, ITME 2016.

Authorized licensed use limited to: Auckland University of Technology. Downloaded on December 22,2020 at 22:33:55 UTC from IEEE Xplore. Restrictions apply.