IMPROVING UZBEK MACHINE TRANSLATION THROUGH PARALLEL CORPORA: CHALLENGES AND SOLUTIONS

Annotasiya

The thesis explores the significance of parallel corpora in modern translation studies, focusing on their crucial role in improving machine translation systems, specifically in the context of the Uzbek language. Parallel corpora, which consist of texts in multiple languages aligned at the sentence or paragraph level, are essential for training neural network-based translation systems. The paper outlines the main challenges in creating high-quality parallel corpora, particularly for underrepresented languages like Uzbek. These challenges include limited available resources, contextual mismatching, errors in segmentation and alignment, and copyright issues. The thesis discusses several solutions to these problems, such as building open-access databases, leveraging machine translation systems, using modern alignment tools, and engaging in crowdsourcing efforts. Additionally, it emphasizes the future potential of parallel corpora in advancing translation quality, supporting linguistic research, and promoting the global recognition of the Uzbek language. Ultimately, the paper argues that parallel corpora are not just a scientific resource but a technological tool, bridging the gap between human translators and machine translation systems.