Public information sources, such as Social Media as well as traditional media, are flooded with content related to the COVID 19 crises. This content covers many types of documents, from factual reports, over opinions, sharing of personal information, disinformation and irrelevant chitchat. Filtering out from this massive content stream relevant factual information requires accurate interpretation of the text content. We report how recent deep learning models for natural language processing can be used to extract facts relevant for health experts for monitoring of early warning signals and trends.