•  
  •  
 

Corresponding Author

Zanbaq Hikmet Thanon

Authors ORCID

Zanbaq Hikmet Thanon: https://orcid.org/0009-0003-3705-4598

Javad Salimi Sartakhti: https://orcid.org/0000-0002-5571-4655

Document Type

Article

Keywords

E-learning, Recommendation system (RS), Topic modeling, Latent dirichlet allocation (LDA), Bag-of-Words (BoW), Coherence score

Abstract

Although the wide range of instructional materials, might overcome students and limit their capability to learn effectively, platform of education become important for providing personalized learning. Traditional recommendation systems (RS) try to resolve this issue; yet, they fail in understanding learning content semantically. This restriction makes it more difficult for them to deliver accurate, useful, and clear recommendations. The aim of this study is to enhance E-learning RS using topic modeling, specially the Latent Dirichlet Allocation (LDA) method, to discover latent structures within educational content. The ``20 Newsgroups'' dataset that containing over 20,000 items across 20 categories used to evaluate the proposed methodology. The training samples applied in this research include 11,314 samples, the test set contains 7,532 samples. The proposed methodology consist of many stages starts with systematic preprocessing (tokenization, stop-word removal, stemming, and lemmatization) also, feature extraction using the Bag-of-Words (BoW) model, and topic modeling through LDA. The evaluation of system's ability to generate meaningful topics using coherence scores and topic comprehension. The experimental results demonstrated that the LDA-based model obtained a coherence score of 0.5399.In addition, the model tested on 191 instances in a Book title category dataset, it achieved a topic relevance score of 0.7736, which made it easy to determine educational materials that are more related. Results of this study indicate that the integration of RS and LDA improves student participation, material detection, and personalization. This work presents a flexible and clear architecture for E-learning RS, highlighting the advantages of merging topic modeling with RS algorithms in order to develop more effective and user-centered digital learning environments.

Download

Share

COinS