Existing studies on social media in the context of crisis have studied the content of items and their patterns of transmission. However, social media content generated during a crisis will generally be unstructured and only reflect the immediate experiences of the authors, while the volumes of data created can make rapid interpretation very challenging. Crisis situations can be characterized with various expected attributes. In many situations there will be large amounts of information relevant to the situation already available. We argue that existing natural language engineering technologies can be integrated with emerging social media content utilization techniques for more powerful exploitation of social media content in crisis response.