Enhancing Personalized News Recommendations with a Hybrid Model: Integrating BERT, Neural Collaborative Filtering and Attention Mechanisms

Suresh, Kiruthika (2024) Enhancing Personalized News Recommendations with a Hybrid Model: Integrating BERT, Neural Collaborative Filtering and Attention Mechanisms. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (821kB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

Personalized news recommendation systems are pivotal in addressing the growing demand for relevant and engaging content. This research explores a hybrid recommendation model combining BERT based content embeddings, Neural Collaborative Filtering (NCF) and an attention mechanism to enhance the accuracy and relevance of news recommendations. The study utilizes the MIND dataset which provides a rich collection of user interactions, news metadata and pre-trained entity embeddings. The proposed model leverages BERT to extract contextual embeddings from news content and NCF to model user-item interactions. An attention mechanism is integrated to prioritize key interactions allowing the system to tailor recommendations based on user behaviour. Through rigorous experimentation, the model is evaluated using metrics such as precision, recall and normalized discounted cumulative gain (NDCG). Findings reveal that the hybrid approach significantly outperforms baseline models, demonstrating improved precision and user satisfaction in recommendations. However, challenges such as overfitting in high dimensional embeddings and computational complexity were identified. Regularization techniques like dropout and L2 regulations were employed to address these issues effectively. The study contributes to the growing body of research in personalized news recommendation systems by integrating advanced NLP and recommendation techniques. Future work will focus on incorporating additional contextual factors example temporal and geographic preferences, exploring lightweight alternatives for real-time deployment. These advancement hold promise for more accurate, scalable and user centric news recommendation systems.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Jilani, Musfira UNSPECIFIED
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science P Language and Literature > P Philology. Linguistics > Computational linguistics. Natural language processing
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	05 Sep 2025 11:25
Last Modified:	05 Sep 2025 11:25
URI:	https://norma.ncirl.ie/id/eprint/8822

Actions (login required)

View Item