Social media such as tweets are emerging as platforms contributing tosituational awareness during disasters. Information shared on Twitter by bothaffected population (e.g., requesting assistance, warning) and those outsidethe impact zone (e.g., providing assistance) would help first responders,decision makers, and the public to understand the situation first-hand.Effective use of such information requires timely selection and analysis oftweets that are relevant to a particular disaster. Even though abundant tweetsare promising as a data source, it is challenging to automatically identifyrelevant messages since tweet are short and unstructured, resulting tounsatisfactory classification performance of conventional learning-basedapproaches. Thus, we propose a simple yet effective algorithm to identifyrelevant messages based on matching keywords and hashtags, and provide acomparison between matching-based and learning-based approaches. To evaluatethe two approaches, we put them into a framework specifically proposed foranalyzing disaster-related tweets. Analysis results on eleven datasets withvarious disaster types show that our technique provides relevant tweets ofhigher quality and more interpretable results of sentiment analysis tasks whencompared to learning approach.
展开▼
