• Proliferation of microblogging and the importance of on-topic tweets for time-critical situational awareness to disaster-affected communities and professional responders
• Challenges
o finding informative and relevant information to accelerate disaster response
o Information overload: Labor-intensive content moderation process as a bottleneck to quickly identify task-related information from incoming messages from the crowd
Auto filtering of relative and informative messages
Automatic categorization of messages
• Research Gap
o Several machine learning techniques have been applied in this domain but all features considered in the classifiers are selected in a data-driven approach without a sound understanding of decision-makers’ information needs and requirements, may result in ineffective filtering and misclassification.
o By incorporating the key words generated by a top-down approach with structured domain-knowledge proposed, the accuracy of classification still has the potentials for improvement