z-logo
open-access-imgOpen Access
Beyond Relevance: Enhancing Serendipity in Content-Based Recommendations with Knowledge Graphs
Author(s) -
Nur Izyan Yasmin Saat,
Masnizah Mohd,
Shahrul Azman Mohd Noah,
Sumaia Mohammed Al-Ghuribi
Publication year - 2025
Publication title -
ieee access
Language(s) - English
Resource type - Magazines
SCImago Journal Rank - 0.587
H-Index - 127
eISSN - 2169-3536
DOI - 10.1109/access.2025.3598342
Subject(s) - aerospace , bioengineering , communication, networking and broadcast technologies , components, circuits, devices and systems , computing and processing , engineered materials, dielectrics and plasmas , engineering profession , fields, waves and electromagnetics , general topics for engineers , geoscience , nuclear engineering , photonics and electrooptics , power, energy and industry applications , robotics and control systems , signal processing and analysis , transportation
Content-based recommender systems suffer from overspecialization, creating filter bubbles that limit user exploration. Serendipitous recommendations that are unexpected and novel yet relevant offer a solution to this problem. Thus, this paper proposes a novel approach to enhance both precision and serendipity in content-based recommender systems using Knowledge Graph (KG) representations by enhancing the distance-based recommendation strategy. Our method combines the MovieLens SAC dataset (which includes details like actors, directors, and genres) with the Movie Plot Synopses with Tags (MPST) dataset to enrich item details. Using Latent Dirichlet Allocation (LDA), we generate topics from plot synopsis and integrate them into the KG alongside other features. User profiles are developed from movies rated ≥3.5 and strengthened using INTERESTED_IN relationships for features with occurrence counts ≥2 in the KG. We implement a two-hop distance approach on the KG to identify items that are semantically related yet sufficiently distant from user preferences. The semantic richness of KGs enables the exploration of hidden relationships between items and user preferences overlooked by conventional systems. Experiments demonstrate that our method significantly outperforms five baseline models (TF-IDF, LDA, r TF-IDF, r LDA, and d TF-IDF) across precision and serendipity measures. Results show our approach successfully discovers novel and unexpected items without sacrificing relevance, addressing the overspecialization problem in traditional content-based recommender systems. Future work will focus on enhancing the graph by integrating it with more contextual data, providing recommendation explanations, and evaluating through real user interactions.

The content you want is available to Zendy users.

Already have an account? Click here to sign in.
Having issues? You can contact us here
Accelerating Research

Address

John Eccles House
Robert Robinson Avenue,
Oxford Science Park, Oxford
OX4 4GP, United Kingdom