An Innovative Approach to Semantic Search in Digital Libraries for Low-Resource Languages of Uzbekistan

Abstract

The 21st century has brought digitization of
knowledge, but as libraries go digital, inclusive search tools are
becoming necessary. Though most digital searches cater to highresource
languages, leaving speakers of low-resource languages,
like Uzbek, underserved. This research proposes an original
concept for semantic searches in digital libraries tailored
around the language, culture, and socio-technical milieu of lowresource
languages like Karapalpak, Tajik, and regional
dialects of Uzbek. Our solution integrates numerous
sophisticated systems to overcome key challenges like the lack
of annotated corpora, embarrassment of morphological
complexity, multilingualism, and others. We design a
comprehensive semantic search engine that provides meaningaware
search results aligned with users' intent, regardless of the
linguistic variations and custom-built domain-specific language
resources, multilingual embeddings, and ontologies used. The
framework enables semi-automatic transliteration, crosslanguage
retrieval, context-driven query expansion, and out-oflanguage
query imposition. The system, implemented as a
prototype on a Uzbekistan digital library corpus,
demonstratively surpasses keyword-based searches in accuracy
and user satisfaction. This research will promote the
development of knowledge systems that are more culturally
aligned and relevant to users, applicable to other languages with
few accessible resources.