Mapping the research landscape of recommender systems for digital libraries

A bibliometric analysis of two decades (2004-2023)

Background of the study: In today's information-rich environment, researchers face the difficulty of managing information overload and struggle to identify meaningful information among a plethora of choices. A recommender system plays a crucial role in assisting users in discovering relevant information from a digital library.

Purpose: This study aims to provide a bibliometric analysis of research publications on recommender systems for digital libraries.

Method: The research method involved the quantitative bibliometric approach to analyse the research publications from the Web of Science and Scopus databases using Biblioshiny. The dataset retrieved comprises 374 documents published between the period of 2004-2023.

Findings: The study's findings highlight that the number of publications was notably high in 2009 and 2018 within the analyzed period. The most significant contributors are Porcel C, Herrera-Viedma E, and Beel J. There is a considerable international collaboration between the countries- China, USA, and Germany. 

Conclusions: This study indicates an emerging interest in recommender systems for digital libraries with the continuous evolution of new recommendation models. There is a huge potential for research considering the availability of multimodal data, the continuous evolution of new technologies such as deep learning, and opportunities for the development of an architecture for integrating different digital libraries.

Introduction

We live in a society that is overwhelmed with an unprecedented amount of information. The rapid expansion of information technologies serves as a catalyst for the explosive growth of digital information(Anas & Salim, 2023). As a result, researchers consistently find themselves engaged in an ongoing battle against information overload and encounter different challenges to discover relevant information precisely when needed or aligned with their preferences. This problem appears especially in the context of the library where the rate of information overload exceeds users’ processing capabilities(Porcel et al., 2018). As a result, traditional library systems have transformed into smart and digital libraries. The application of digital libraries extends across various contexts; however, our study specifically concentrates on their usage within an academic environment. University Digital Libraries (UDL) surpasses traditional libraries by providing a variety of services and resources, such as digital collections, online databases, and multimedia data to meet the different needs of students, faculty, and staff in their learning, teaching, and research endeavors(Al-Qallaf & Ridha, 2019);(Kato et al., 2021).

This constant fight against information overload requires the development of effective strategies and tools. The Recommender Systems (RS) plays a crucial role in tailoring information to individual research interests, aiding in the battle of overwhelming data by providing personalized recommendations(Adomavicius et al., 2021);(Burke et al., 2015). This system assists users in navigating relevant information based on their past preferences and provides personalized suggestions for various library resources. These recommendations encompass a wide range of recommendations including book(Mo et al., 2019-08), research article/paper(Magara et al., 2017-12), journal(Ghosal et al., 2019-06), audio(Roy et al., 2020), and many more within digital libraries.

Various types of RS have been implemented within digital libraries based on fuzzy linguistic modelling(Tejeda-Lorente et al., 2014), collaborative filtering(Shen, 2018);(Nugraha et al., 2020), and hybrid approaches(Amolochitis, 2018). The users of digital libraries have varied interests and to deal with these unpredictable needs, some attempts have been made to model and understand user behavior through implicit user activities to improve recommendations for digital library services(Akbar et al., 2014-09);(Liu, 2022). However, the quality of recommended items remains a challenge in many existing approaches.(Tejeda-Lorente et al., 2014)addressed this gap by proposing a novel recommender system based on item quality, incorporating a fuzzy linguistic approach, and testing it within a UDL to enhance users' access to relevant research resources. Despite the progress made, personalized recommendations in digital libraries still face challenges such as cold start problems. To address this issue,(Tejeda-Lorente et al., 2018)incorporated a fuzzy linguistic approach utilizing bibliometrics to reduce the necessity for user interaction.(Porcel et al., 2017)examined various proposals for RS to improve information access, with a focus on identifying the best fuzzy linguistic modeling RS to assist users in academic library services.

Recent advancements have been made in RS within digital libraries based on user preferences utilizing the Latent Dirichlet Allocation model to recommend items(Zhao, 2021), ontology-based multi-attribute collaborative filtering RS to enhance the access to learning resources within digital libraries(Senthil Kumaran & Latha, 2023).(Liu, 2021-09)proposed a personalized RS using SOM neural networks to enhance the web access behavior of users for UDL. Despite the ongoing development of various RS in the field of digital libraries, the existing literature remains somewhat limited. This field is still in its nascent stage, and there are very few articles, especially in UDL. Thus, we consider the landscape of the digital library to take into account the potential for the development of a unified RS model for digital libraries.

To the best of our knowledge, nevertheless, so far, no research has been conducted to study the bibliometric analysis of RS for digital libraries. The measurement of research using bibliometrics typically focuses on topics, scientific fields, and publications(Muntiah & Dewi, 2023). Our research addresses the gap by analyzing previous studies, offering important insights, and highlighting potential research opportunities within the field.

The objective of our study is to provide an overview of the field’s current state, progress, and influence through bibliometric analysis. The extensive datasets were obtained from the Web of Science (WoS) and Scopus covering a span of two decades (2004-2023). The analysis not only investigates publication trends and collaboration patterns within the field but also underscores the need for a unified RS framework for digital libraries. This need is highlighted, especially in the context of UDL, where the current landscape displays a diverse spectrum of RS implementations. To confront this inevitable challenge, this study aims to provide valuable insights that have the potential to shape the development and standardization of RS by addressing the unique requirements and complexities of academic environments within the UDL. This research would facilitate further research endeavors. The following research questions to ensure the continuity of RS for digital libraries:

RQ1. How has the annual scientific publication growth in RS for digital libraries evolved over the past two decades (2004-2023)?

RQ2. Who are the most prolific authors in this relevant field?

RQ3. What is the structure of the co-citation network between authors working in this field over the past two decades (2004-2023)?

RQ4. Which countries make the largest contributions based on publications and citations in this field?

RQ5. Which keywords are the most frequently