Towards Effective Named Entity Recognition in Uzbek Medical Contexts

In this research work, the authors developed an algorithm for recognizing named entities adapted for medical texts in the Uzbek language. The Python's Spacy library was chosen as the main technology for implementing the algorithm, including its multilingual model for training its own (custom) model for the Uzbek language. The researchers also conducted a comparative analysis of similar works devoted to similar problems and issues, and provided objective reasons for the relevance of the proposed work. In addition, the authors also tested the algorithm with different datasets in order to identify the effectiveness of detecting named entities, including medical terminology. As a result of testing the algorithm on 1000 sentences, the average accuracy rate reached 92%, which can be indicated as a good efficiency. In the final part, the authors note possible directions for the development of the work, in particular, the use of alternative neural network architectures and a marked corpus.