International Journal of Information Science and Management (IJISM)

Abstract The rapid evolution of deep learning (DL) and the exponential growth of digital collections have opened transformative opportunities for library and information services. While prior studies have explored artificial intelligence and machine learning applications in libraries, a focused synthesis of deep learning and neural network applications, their dominant use cases, and associated challenges in library and information science (LIS) remains fragmented. Following the PRISMA 2020 guidelines, 45 peer-reviewed articles indexed in Scopus (1998–2024) were systematically selected and analyzed after screening 246 initial records. Recommendation systems emerged as the dominant application (31%, n=14), followed by library and digitization management, text mining and natural language processing (NLP), user behavior prediction, and image/video processing. The most frequently employed architectures were Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), Transformers (BERT/GPT variants), Graph Neural Networks (GNNs), and Gated Recurrent Units (GRUs). Despite significant performance gains, adoption remains hindered by high implementation costs, insufficient staff expertise, poor data quality, inadequate infrastructure, and limited access to advanced tools. Overcoming these challenges requires AI literacy programs, scalable IT infrastructure, improved data quality, and cross-institutional collaboration, enabling adaptive, data-driven library services and reinforcing libraries' evolving role in the digital age.