Providing recommendations for communities of learners in MOOCs ecosystems

https://doi.org/10.1016/j.eswa.2022.117510Get rights and content

Highlights

  • A content-based recommendation using topic modeling and labeling techniques.

  • Recommendation of part of courses from multiple providers to support learners.

  • Implementation of a system with real data from providers, enabling the expansion.

  • The model’s number of topics and document labels are automatically defined.

  • Three quali-quantitative experiments confirm the system’s effectiveness.

Abstract

Massive Open Online Courses (MOOCs) have been widely disseminated due to the arrival of Web 2.0. However, the growth of MOOCs brings some difficulties for students in choosing suitable courses in these ecosystems. In recent years, some recommendation systems emerged to solve this problem but remain limited since they do not identify the student’s prior knowledge broadly or the student’s goals. To overcome this limitation, this work proposes the Fragmented Recommendation for MOOCs Ecosystems (FReME), a recommendation system to suggest parts of courses from multiple providers (i.e., Khan Academy, Udemy, and edX). FReME is based on the student profile and on the MOOCs ecosystems perspective to balance the ecological environment and strengthen interactions. Moreover, we differ from the current recommendation systems since our method identifies and reduces the students’ knowledge gap optimizing the learning process. Experimental results conducted with a dataset integrating 3 MOOCs providers and 19 students demonstrated that the implemented techniques are more consistent than other approaches. Finally, it was verified through precision, utility, novelty, and confidence that our recommendations are 62,24% accurate, 68.89% useful, 72.81% reliable, and present new content in 99.12% of cases. These results validate that FReME supports students in reducing their knowledge gap.

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)

  • BarrosH. et al.

    Linked educational online courses to provide personalized learning

  • BasiliV.R.

    Software modeling and measurement: The goal/question/metric paradigmTechnical Report

    (1992)
  • BhattC. et al.

    SeqSense: Video recommendation using topic sequence mining

  • BiagiottiB.A.

    Assessment of the quality of information from massive courses: A case study of TELELAB

    (2016)
  • BoschJ.

    From software product lines to software ecosystems

  • CamposR.

    Recommendation system for knowledge acquisition in MOOCs ecosystems

    (2019)
  • CamposR. et al.

    Web-based recommendation system architecture for knowledge reuse in MOOCs ecosystems

  • CamposR. et al.

    A recommendation system based on knowledge gap identification in moocs ecosystems

  • CamposR. et al.

    Recommendation system for knowledge acquisition in MOOCs ecosystems

  • CamposR. et al.

    Supporting students through a recommendation system for knowledge acquisition in MOOCs ecosystems

  • CamposR. et al.

    A recommendation system enhanced by topic modeling for knowledge reuse in MOOCs ecosystems

  • CamposR. et al.

    Techniques, challenges and future directions of recommendation systems for MOOCs

    IEEE Transactions on Learning Technologies

    (2021)
  • Cited by (3)

    View full text