Smart Semantic Repositories for Localized Library Discovery Systems

Thesis Abstract

This study investigates how smart semantic repositories can enhance localized library discovery systems to support precise, context-aware access to multilingual and multimedia library resources in urban public libraries. The problem addressed is the fragmentation of heterogeneous metadata and the limited semantic interoperability across local catalogues, which impedes users’ ability to locate relevant items efficiently in culturally and linguistically diverse settings. The aim is to design, implement, and evaluate an ontology-driven, machine-actionable repository that integrates autonomous reasoners, linked data principles, and user-context modeling to deliver superior discovery accuracy and personalization. Specific objectives include (1) developing a localized ontology for library resources that captures genre, format, language, authorial networks, and user profile signals; (2) implementing a semantic indexing and mapping pipeline leveraging schema.org, BIBO, SKOS, and domain-specific extensions to normalize heterogeneous metadata; (3) integrating a user-context engine that dynamically adapts search results based on locale, user history, and device; (4) evaluating retrieval effectiveness against baseline catalogs using precision, recall, and F-measure across 12 diverse catalogues; and (5) assessing user satisfaction and discovery efficiency through mixed-methods analysis. The methodology adopts a positivist-constructivist hybrid research design combining prototypical development with empirical evaluation. The population comprises 15 urban public libraries across three districts, with a total collection of approximately 2.5 million items and 500,000 user queries collected over six months. A stratified random sample of 10 libraries is selected for pilot deployment, and within each library a sample of 1,200 user sessions is recorded (total n ? 12,000 sessions). Data collection instruments include (i) system logs capturing query terms, click streams, and dwell time; (ii) an expert-validated ontology development rubric; (iii) standardized usability questionnaires employing SUS and the System Usefulness Scale; and (iv) semi-structured interviews with 30 librarians and 60 frequent patrons. Data analysis employs a combination of quantitative and qualitative techniques statistical comparison of retrieval performance using paired t-tests and ANOVA to identify improvements over baseline catalogs, precision-recall curves, and effect sizes; regression analyses to determine predictors of user satisfaction and effective discovery; and thematic analysis of interview transcripts to interpret perceived usefulness and contextual fit. The analytical framework also includes semantic similarity metrics, knowledge graph reasoning to evaluate inference paths, and a pilot-based A/B testing scheme to compare the semantic repository against a traditional metadata-driven system. Validity and reliability are addressed through triangulation, inter-rater reliability for ontology annotation (Cohen’s kappa ? 0.80), and pilot testing with 1000 random queries for stability. Expected findings indicate that the smart semantic repository will produce statistically significant improvements in precision and recall (expected average F1 improvement of 15–22% over baseline catalogs), higher user task success rates, and reduced time-to-first-relevant-click by approximately 18–25%. The user-context engine is anticipated to yield context-aware rank-adjusted results that align closely with local linguistic preferences and device constraints, particularly enhancing retrieval for non-dominant languages and specialized formats such as digital manuscripts and audio-visual materials. The study is also expected to reveal key patterns in librarian workflows, illustrating how semantic augmentation reduces cataloguing bottlenecks and supports more accurate metadata creation. Contributions to knowledge include (1) a replicable, locale-aware ontology and linked-data model for library discovery that bridges heterogeneous metadata schemas; (2) an end-to-end semantic indexing and reasoning pipeline tailored for localized collections; (3) empirical evidence on user-centered design and impact on discovery efficiency in public libraries; and (4) a methodological framework for evaluating semantic repositories in real-world library settings. The main conclusion is that smart semantic repositories, when grounded in robust ontologies and user-context modeling, significantly enhance localized library discovery by improving access to multilingual and multimedia resources and by aligning search results with user intent and local needs. Practical recommendations include adopting open standards for metadata interoperability, investing in librarian training for ontology curation, and scaling the prototype to regional networks to sustain semantic interoperability across districts. Future research directions entail extending the framework to include multimodal search capabilities, cross-institutional authentication, and adaptive ranking mechanisms informed by evolving user behavior.