Tornadoes, hurricanes, earthquakes, plane crashes, bombings, terrorist attacks - deadly disasters happen very frequently. Social media platforms such as Twitter have emerged as an effective tool for broadcasting messages worldwide during disaster events. However, with millions of messages posted through these services during such events, it has become very difficult to find useful information. Tweet hashtags or keywords have the potential to improve the search for information, but unfortunately, there is a large number of disaster-related tweets that do not have any user-provided hashtags or they are not useful for disaster response. In this talk, Cornelia Caragea shares how deep learning models can automatically extract useful information in the form of hashtags or keywords that can help emergency response organizations to become more situationally aware during and following a disaster.