AI Transforms Library Search from Keywords to Knowledge

Libraries are evolving from static repositories into dynamic knowledge hubs, but their search systems often lag behind. Traditional Online Public Access Catalogues (OPACs) rely on keyword indexing and Boolean queries, which struggle with the rapid growth of scholarly literature, leading to inefficient and user frustration. This gap highlights a pressing need for intelligent systems that understand user intent and semantic relationships. A new study addresses this by proposing a Smart OPAC framework that transforms conventional catalogues into AI-driven tools, leveraging knowledge graphs to enhance relevance and exploration. This approach could significantly improve how researchers and students navigate digital collections, making information retrieval more intuitive and effective.

The researchers developed a Smart OPAC that integrates artificial intelligence and knowledge graph techniques to enable semantic search, thematic filtering, and interactive visualization. Unlike traditional OPACs, this system retrieves scholarly data in real-time from multiple sources—Europe PMC, OpenAlex, and Semantic Scholar—using their APIs. It employs SBERT (Sentence-BERT) to create dense vector representations of paper titles, moving beyond simple keyword matching to capture semantic meaning. KeyBERT is used for keyword extraction, identifying themes from documents based on cosine similarity scores. These themes are then presented as dynamic multi-select filters, allowing users to narrow interactively, with papers retained only if their extracted themes match selected concepts.

Quantitative evaluation demonstrated substantial improvements in retrieval efficiency and relevance. The system showed fast average retrieval times: Semantic Scholar at 1.124 seconds and Europe PMC at 1.318 seconds, though OpenAlex was slower at 3.646 seconds. Relevance effectiveness varied by query, with well-defined topics like "semantic search libraries" achieving 83.33% relevance after filtering, while broader queries like "AI in OPAC" dropped to 30%. Information overload reduction was significant, with reductions up to 90% for complex queries such as "AI in OPAC" and "Information Retrieval LIS" when using Europe PMC, as shown in Table 4 and Figure 3. Source contribution analysis revealed a balanced distribution, with Europe PMC and OpenAlex each contributing 45.45% of retrieved papers, and Semantic Scholar adding 9.09%, supporting the use of heterogeneous sources for comprehensive coverage.

Of this research are profound for modernizing digital library services. By reducing information overload and enhancing relevance, the Smart OPAC framework supports next-generation research workflows, enabling users to explore interdisciplinary connections more effectively. Knowledge graph visualizations, as illustrated in Figures 4, 5, and 6, provide qualitative insights by mapping semantic relationships and thematic clusters, moving beyond list-based interfaces to foster deeper intellectual engagement. This system addresses the limitations of traditional OPACs by offering context-sensitive access, which could benefit academic institutions, public libraries, and any organization managing large digital collections, ultimately making scholarly more accessible and efficient.

However, the study has limitations that warrant further investigation. The evaluation was based on a limited number of queries and data sources, and user-centered usability tests were not included, which may affect generalizability. Future work aims to expand data source coverage, incorporate user interaction analytics, and implement personalization based on feedback. Additionally, dynamic ontology enrichment and real-time knowledge graph updates are planned to enhance the system's adaptability. These steps are crucial for ensuring the Smart OPAC can evolve into a robust tool for advanced research and decision-making in digital library environments, addressing current gaps while paving the way for more intelligent systems.