A DOMAIN-KNOWLEDGE-BASED MODEL FOR REFERENCE RESOLUTION IN LOW-RESOURCE LANGUAGES

Abstract

Reference resolution is a fundamental task in Natural Language Processing (NLP), essential for achieving coherent text understanding. However, state-of-the-art deep learning approaches heavily rely on large-scale annotated corpora, posing significant challenges for low-resource languages where such data is scarce. This paper proposes a domain-knowledge-based model designed to improve reference (coreference) resolution in low-resource settings. The proposed hybrid approach integrates statistical learning with explicit domain knowledge derived from ontologies and semantic constraints. By leveraging external knowledge bases to guide the resolution process, the model compensates for the lack of sufficient training data. Experimental results demonstrate that incorporating domain knowledge significantly outperforms baseline statistical models in terms of accuracy and F1-score. This research contributes to the development of more robust information extraction and text analysis systems for languages with limited digital resources.