The ever-increasing and publicly accessible flow of news represents a valuable resource for extracting information to detect health crises using machine learning techniques. The primary objective of this thesis is to accurately classify news articles by topic, with a focus on those related to health, and subsequently categorise them by specific diseases. This research aims to explore innovative architectures and techniques in Natural Language Processing to analyse and classify the news stream. In addition, a robust test suite will be developed to evaluate the reliability of the developed models. The resulting classifications will generate time series of news related to specific diseases.