Situation awareness plays an essential role in making real-time decisions in mass gatherings. In the last few years, social media data analysis has been proved to be an effective approach to enable and enhance situation awareness. Mass gathering events are dynamic and critical environments where thousands of people attend. During the event, there is a potential for injuries and other health hazards, and thus it is critical for emergency medical services to access real-time and situational awareness information, especially concerning the nature of the crowd. It has been well recognized in the literature that crowd mood and behaviour can have a direct impact on the crowd safety and patient presentation rates. We describe a mobile social media-enabled crowd monitoring architecture that aims to improve emergency management decision-making by analysing the data from social networks in real-time. The proposed architecture incorporates a crowd behaviour classification model, which facilitates real-time situation awareness and provides a better understanding of analysis results. Awareness and perception of crowd mood and behaviour during the event can significantly improve prediction of patient presentation rates; leading to timely and effective medical care provision. The implementation and evaluation of the proposed framework on an Android mobile phone is described.