Providing recommendations for communities of learners in MOOCs ecosystems
Introduction
With the development of the Internet, open online education has been widely used for bringing new values based on ethics of participation, openness, and collaboration, which makes it a facilitator for the issues faced by traditional learning methods (Peters, 2008). One of the modalities of this movement is the Massive Open Online Courses (MOOCs) that are courses open to any student, anywhere, at almost any time, and with a very large number of students enrolled. MOOCs are run by MOOCs platforms, which can be proprietary (e.g., Udacity) or open-source (e.g., Open edX) (Kim, 2014).
In the context of e-learning in general, some authors point to the benefits of exploring distance education platforms based on the software ecosystem perspective, which is defined as a set of businesses operating as a unit that interacts with a shared market for software and services, with relationships among them underpinned by a common market or technological platform (Jansen et al., 2009). It becomes appropriate due to the demand for inter-organizational reuse (Bosch, 2009), either from learning objects repositories or even from educational services, such as forums and chats. Therefore, some work points out characteristics, roles, or definitions of software ecosystems to be applied specifically in the context of MOOCs, i.e., to analyze MOOCs in a broader perspective so-called MOOCs ecosystems (Campos et al., 2018).
Due to the COVID-19 pandemic, the number of courses and users in MOOCs has been growing worldwide. In the year of 2020, the top-3 MOOCs providers (i.e., Coursera, edX, and FutureLearn) had 2800 new courses with more registered users in April 2020 than in the whole year of 2019 (Shah, 2020a, 2020b). The pandemic caused a spike in unemployment and accelerated the demand for professionals trained in the jobs of the future, composed of new technologies, sectors, and markets in more interconnected global economic systems (World Economic Forum, 2020). There was an increase in demand for online learning, with employed people seeking out more personal development courses and unemployed people with a larger emphasis on digital learning skills (World Economic Forum, 2020). In this scenario, besides offering these most sought courses, MOOCs play a relevant role in democratizing learning without geographic boundaries and with several free courses provided by universities around the world (Shah, 2020b).
However, with the growth of MOOCs platforms, problems arose for students, as the difficulty in finding the most suitable course that matches their expectations or needs. There are countless possibilities of courses distributed at different providers which can be a positive aspect. However, in many cases, it causes information overload with students without the proper guidance on choosing the most appropriate content. Previous research has suggested that more than half of dropouts in MOOCs may occur when the enrolled course was not the one that students were seeking or when their expectations were not attended by the course (Zhang et al., 2021).
Therefore, in this work, we propose a solution to this problem based on an investigation of software ecosystem strategies in such a scenario considering the following main research question (RQ): How to identify and reduce knowledge gaps in the MOOCs ecosystems?
Considering that there is a gap between the knowledge already obtained by the student and the new knowledge required to reach the desired expertise, this work aims at supporting in reducing specifically such users’ knowledge gaps, i.e., the knowledge that the student does not yet have about a particular subject of interest. To do so, we combine recommendation systems with users’ data from MOOCs providers and propose the Fragmented Recommendation for MOOCs Ecosystems (FReME) to investigate how to assist them in achieving their goals based on the recommendation of parts of courses (i.e., courses’ modules),1 considering the perspective of MOOCs ecosystems. These recommendations apply for full courses if providers do not organize content into modules, for example.
Several recommendation systems have been proposed to support MOOCs adopting different approaches, such as Collaborative Filtering (CF) or content-based recommendation. However, these efforts may fail to identify the most adequate knowledge broadly, since they focus on specific MOOCs providers, or recommend complete courses, or videos from MOOCs platforms. Thus, it is necessary to propose solutions that consider student’s profiles using multiple platforms and fragmented knowledge aiming at the reduction of their knowledge gap. The learning item fragmentation is a target of future work in recent studies applied to recommendation in MOOCs, remaining a challenge in this field. Jing and Tang (2017) ratify the need to improve recommendations in MOOCs not only recommending full courses, but also “different kinds of content such as video and knowledge” since it can improve the personalization level of recommendations in MOOCs. Barros et al. (2018) emphasize plans to include “the recommendation of a set of topics of a course for a student, not just a full course, in order to provide more accurate recommendations”.
This article is organized, as follows: Section 2 describes the main concepts of our work. Section 3 reviews the related work. Section 4 introduces the FReME conceptual model, details the content-based and the topic labeling methods, and presents a example of recommendation. Section 5 presents our experimental results, and this article is concluded in Section 6.
Section snippets
Background
In this section, we describe the main concepts addressed in this paper: recommendation system and MOOCs ecosystems.
Related work
In our previous work, we conducted a Systematic Mapping Study (SMS) (Campos, 2019, Campos et al., 2021) to analyze studies that present recommendation systems applied to MOOCs. The most used approach in this context is CF, such as in the study of Symeonidis and Malakoudis (2019). However, this approach is only feasible in scenarios where it is possible to extract data from several users, which is not the case of the present work since it aggregates several providers including those that only
The proposed system
From the MOOCs ecosystems definition, identifying the actors’ roles and their relationships (Campos et al., 2018), it was possible to propose the FReME conceptual model, as shown in Fig. 2. Given the restriction of some providers regarding the extraction of private data from users, the system contains an authorization layer (3) that is activated as soon as the user (1) requests a recommendation from the system (2). Data (5) from that user (i.e., authorized data from MOOCs ecosystems) is
Evaluation and results
This section presents an evaluation method based on both qualitative and quantitative analyses that is composed of a set of three experiments conducted to verify our work and validate its implemented techniques, i.e., the topic modeling, the topic labeling, and the recommendation method itself based on real data from multiple providers. Therefore, the first experiment compares the effectiveness of our improved topic modeling technique with a widely used topic modeling technique - LDA. The
Conclusion
This work investigated the problem faced by students in MOOCs in choosing the best course due to the increase in the number of courses on these platforms. It focused on answering the main research question RQ: How to identify and reduce knowledge gaps in the MOOCs ecosystems? Overall, we proposed a new content-based recommendation system applied to the scenario of MOOCs - FReME.
CRediT authorship contribution statement
Rodrigo Campos: Methodology, Software, Investigation, Resources, Writing – original draft, Visualization. Rodrigo Pereira dos Santos: Validation, Writing – review & editing, Supervision. Jonice Oliveira: Conceptualization, Validation, Writing – review & editing, Supervision.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
We thank IFRJ, Brazil for supporting this research. We also thank CAPES, Brazil, CNPq, Brazil, and FAPERJ, Brazil (Proc. 211.583/2019) for their financial support, and Oracle Cloud credits and related resources provided by the Oracle for Research program (award number CPQ-2160239).
References (45)
- et al.
MOOCS and serials
Serials Review
(2013) - et al.
Persisting big-data: The NoSQL landscape
Information Systems
(2017) - et al.
CARM: Confidence-aware recommender model via review representation learning and historical rating behavior in the online platforms
Neurocomputing
(2021) - et al.
An analysis of the coherence of descriptors in topic modeling
Expert Systems with Applications
(2015) - et al.
Deep variational matrix factorization with knowledge embedding for recommendation system
IEEE Transactions on Knowledge and Data Engineering
(2019) - et al.
Multi-modal matrix factorization with side information for recommending massive open online courses
Expert Systems with Applications
(2019) - et al.
The learning behaviours of dropouts in MOOCs: A collective attention network perspective
Computers and Education
(2021) Recommender systems
(2016)- et al.
Online courses recommendation based on LDA
- et al.
MoocRec: Learning styles-oriented MOOC recommender and search engine
Linked educational online courses to provide personalized learning
Software modeling and measurement: The goal/question/metric paradigmTechnical Report
SeqSense: Video recommendation using topic sequence mining
Assessment of the quality of information from massive courses: A case study of TELELAB
From software product lines to software ecosystems
Recommendation system for knowledge acquisition in MOOCs ecosystems
Web-based recommendation system architecture for knowledge reuse in MOOCs ecosystems
A recommendation system based on knowledge gap identification in moocs ecosystems
Recommendation system for knowledge acquisition in MOOCs ecosystems
Supporting students through a recommendation system for knowledge acquisition in MOOCs ecosystems
A recommendation system enhanced by topic modeling for knowledge reuse in MOOCs ecosystems
Techniques, challenges and future directions of recommendation systems for MOOCs
IEEE Transactions on Learning Technologies
Cited by (3)
Content recommendation with two-level TransE predictors and interaction-aware embedding enhancement: An information seeking behavior perspective
2023, Information Processing and ManagementA Systematic Literature Review on the Implications of Educational Recommender System in Teaching Learning Environment
2023, International Journal of Engineering Trends and TechnologyPredictors of continued intention of working professionals for pursuing e-learning courses for career advancement
2023, Information Discovery and Delivery