A Novel Recommendation of Learning Items Based on Deep Neural Networks and Trust Relationships
Saved in:
| Title: | A Novel Recommendation of Learning Items Based on Deep Neural Networks and Trust Relationships |
|---|---|
| Language: | English |
| Authors: | Yamina Aissaoui, Lamia Berkani, Faiçal Azouaou |
| Source: | International Journal of Learning Technology. 2024 19(4):397-421. |
| Availability: | Inderscience Publishers. World Trade Centre Building II 29 route de Pre-Bois Case Postale 856 CH-1215, Geneva 15, Switzerland. e-mail: editor@inderscience.com; Web site: https://www.inderscience.com/jhome.php?jcode=ijlc |
| Peer Reviewed: | Y |
| Page Count: | 25 |
| Publication Date: | 2024 |
| Document Type: | Journal Articles Reports - Research |
| Descriptors: | Information Networks, Online Searching, Search Strategies, User Needs (Information), Social Networks, Interest Inventories, Profiles, Open Educational Resources, Trust (Psychology), Electronic Learning, Information Retrieval |
| DOI: | 10.1504/IJLT.2024.143969 |
| ISSN: | 1477-8386 1741-8119 |
| Abstract: | With the rapid development of information technologies, online learning platforms have become the most convenient way for users(teachers and learners) to share their learning content. However, due to the increasing number of educational content, it becomes very difficult for learners to find the most appropriate items. Most previous methods focused on users' ratings to establish learner's profiles, while recent work has added users' comments. In this work we are interested on the usage of social learning networks and we propose a novel learning items recommendation approach through sentiment analysis. Two different deep neural network models (DNNs) have been used, namely: bidirectional encoder representations from transformers (BERT) and recurrent neural networks (RNNs) with their modified version, long short-term memory (LSTM). These models are based on the learners' data, including their favourite content and comments. To support learners in selecting learning resources, a list of trusted learners was developed using similarity between learners. In order to evaluate the effectiveness of our proposal, experiments have been conducted on two different datasets. The results we have obtained demonstrated that our approach outperforms the baselines and related work. |
| Abstractor: | As Provided |
| Entry Date: | 2025 |
| Accession Number: | EJ1458116 |
| Database: | ERIC |
| Abstract: | With the rapid development of information technologies, online learning platforms have become the most convenient way for users(teachers and learners) to share their learning content. However, due to the increasing number of educational content, it becomes very difficult for learners to find the most appropriate items. Most previous methods focused on users' ratings to establish learner's profiles, while recent work has added users' comments. In this work we are interested on the usage of social learning networks and we propose a novel learning items recommendation approach through sentiment analysis. Two different deep neural network models (DNNs) have been used, namely: bidirectional encoder representations from transformers (BERT) and recurrent neural networks (RNNs) with their modified version, long short-term memory (LSTM). These models are based on the learners' data, including their favourite content and comments. To support learners in selecting learning resources, a list of trusted learners was developed using similarity between learners. In order to evaluate the effectiveness of our proposal, experiments have been conducted on two different datasets. The results we have obtained demonstrated that our approach outperforms the baselines and related work. |
|---|---|
| ISSN: | 1477-8386 1741-8119 |
| DOI: | 10.1504/IJLT.2024.143969 |
