Artificial Intelligence–Driven Knowledge Management in Academic Libraries: A Framework for Enhancing Information Retrieval Efficiency and User-Centred Service Delivery

Abstract

The expansion of digital resources in academic libraries (AL) has intensified the need for effective information retrieval systems and services that align with users’ needs. This study introduces an Artificial Intelligence-Driven Knowledge Management Framework for AL, designed to optimise retrieval processes while supporting user-focused service delivery. The framework combines Named Entity Recognition (NER) with Word Sense Disambiguation (WSD) to interpret both the semantic content and underlying intent of user queries. A voting-based NER method, built on Intelligent Monarch Butterfly Optimised Residual Bidirectional Gated Recurrent Unit (IMB-Res-BiGRU) classifiers, ensures reliable entity identification in English and Chinese texts, while an example-driven WSD technique addresses semantic ambiguities to enhance query interpretation. The framework was tested on samples drawn from English and Chinese collections, following text cleaning and normalisation procedures. For document retrieval, Bidirectional Encoder Representations from Transformers (BERT) was applied to achieve high relevance and efficiency. User-oriented service provision was accomplished by tailoring search outcomes according to contextual factors, user preferences, and information-seeking patterns, enabling students, researchers, and faculty to access precise and meaningful resources. Implementation in Python produced experimental outcomes with high accuracy, achieving 97.8% precision and 98.2% recall, thereby confirming significant advances in retrieval effectiveness alongside improved user-focused service delivery. This AI-driven framework demonstrates a scalable model for contemporary AL, uniting advanced knowledge management with personalised services to strengthen access to information and enhance user satisfaction.