Medical Terminology Extraction Using Hybrid Approach for Uzbek Texts

The article presents a hybrid approach to the task of extracting medical terminology from Uzbek-language texts. The developed algorithm combines the capabilities of neural networks and text preprocessing rules, including the conversion of Cyrillic characters to Latin and standardization of dialect forms. A specialized medical corpus of texts was created, covering three different sources: patient diagnoses, medical prescriptions, and medical news. The study identified five main categories of entities: persons, organizations, locations, names of diseases and drugs. The testing confirmed the effectiveness of the proposed approach and revealed the strengths and limitations of the model when analyzing texts of various formats and styles. The results obtained allow us to recommend this algorithm for practical use in digital medicine systems of the Republic of Uzbekistan. In addition to this, there is quite deep comparative analysis of the existing scientific works, which additionally demonstrates actuality of the research. Moreover, authors also added necessary information about Uzbek language to make it easier understand about nature of the language.