This dissertation asks whether NLP-based semantic message prioritisation improves how urgent disaster messages are handled in opportunistic emergency communication, using the 2015 Nepal earthquake as a case study. The short answer: better priority classification helps most when transmission and storage resources are contested, but it cannot overcome the mobility and contact constraints of opportunistic networks.