Mobile devices and microblogging are emerging as significant means of spreading information during disasters not only to affected communities but also to professional responders for time-critical situational awareness (Imran et al., 2013). As the usefulness of information varies significantly, it is common for human annotators to examine the content of incoming messages and control their dissemination in a content moderation process (Link et al., 2013). The large volume, high velocity, and variety of user-generated contents from both mobile devices and social media platforms such as Twitter create a processing bottleneck in finding relevant and valuable information.
With on-topic tweets being important elements for time-critical situational awareness, the affected participants together with the disaster response team can gain from a system that automatically classifies relevant information from Twitter depending on whether or not they increase the value of situational awareness. Therefore, the detection of informative tweets should be one of the main system characteristics where the final output from must consist of brief and self-contained pieces of information that most likely augment situational awareness (Link et al., 2013).
Machine learning plays a vital role in classifying structured information accurately from the amorphous text-based microblogging messages (Imran et al., 2014). Analyzing evolving issues and their emergent effects are somewhat intricate because, even though the classification of information through machine learning is possible, most of the SML classifiers only focus on binary classification e.g. AIDR and Tweedr (Ashktorab et al., 2014). Emphasis on discrete data alone is not adequate to spawn valuable intuitions that control stakeholder response. The problem of multiclass classification in the categorization of messages results in the variance of classifiers’ performance between categories e.g. AIDR and EMERSE.
Different message ontologies exist for different interests of different organizations and, as thus, it is preferable to have SML-based classifiers with minimal variance among the categories since the state-of-the-art machine learning techniques can help to classify messages into sets of fine-grained classes. The current SML-based classifiers do not consider domain knowledge the result of which is ineffective filtering and classification, but, the utilization of new Twitter abilities can aid crisis managers to evaluate the relationships of discrete data so as better comprehend the evolving issues and their emergent effects (Imran et al., 2013).
Disastrous situations pose unique challenges to researchers conducting a study about them prompting researchers to call for particular methods of research (Stallings, 2002). Sudden onset crises present situations with rampant cases of uncertainties that call for urgent decision-making processes and in most cases with inadequate information. Research shows that social media communications, such as Twitter, play an important role in reducing such inadequacy of information particularly in times of disaster. A case in point was the Haiti Earthquake disaster where Twitter’s ability to spontaneously classify tweets and messages conveyed important information to the proper workforces in a well-timed and resourceful manner to address the most critical needs (Caragea et al., 2011).
At present, the extensive embracing of social media in times of disasters creates opportunities for broadcasting critical information that would otherwise be unavailable and in such times, the Emergency Response Units regularly avail such information inform of alerts. In contrast, social media enables a top-down mode of communication. The information available in Twitter is a reflection of people’s experiences and witnesses (Hughes and Palen, 2009) and, as such, allows the affected parties and those offering assistance to gain first-hand insight regarding the situation in near real-time. The modern social media tools have robust technologies to generate actionable knowledge and support decision-making processes with timely analytical insights (Yin e al., 2012).
Ashktorab, Z., Brown, C., Nandi, M., & Culotta, A. (2014). Tweedr: Mining Twitter to Inform Disaster Response. 11th International Conference on Information Systems for Crisis Response and Management.
Caragea, C., McNeese, N., Jaiswal, A., Traylor, G., Kim, H., Tapia, A., . . . Yen, J. (2011). Classifying Text Messages for the Haiti Earthquake. Proceedings of the 8th International ISCRAM Conference.
Imran, M., Elbassuoni, S., Castillo, C., Diaz, F., & Meier, P. (2013). Extracting Information Nuggets from Disaster- Related Messages in Social Media. 791–800.
Imran, M., Castillo, C., Diaz, F., & Vieweg, S. (2014). Processing Social Media Messages in Mass Emergency: A Survey. 4.
Imran, M., Castillo, C., Lucas, J., Meier, P., & Rogstadius, J. (2014). Coordinating Human and Machine Intelligence to Classify Microblog Communications in Crises. In Proceedings of the Companion Publication of the 23rd International Conference on World Wide Web Companion, 159-162.
Link, D., Hellingrath, B., & De Groeve, T. (2013). Twitter Integration and Content Moderation in GDACSmobile. Proceedings of the 10th International ISCRAM Conference, 67–71.-67–71.
Owoputi, O., O’Connor, B., Dyer, C., Gimpely, K., Schneider, N., & Smith, N. (2013). Improved Part-of-Speech Tagging for Online Conversational Text with Word Clusters. Proceedings of NAACL-HLT 2013, 380–390-380–390.
Peffers, K., Tuunanen, T., Rothenberger, M., & Chatterjee, S. (2007). A Design Science Research Methodology for Information Systems Research. Journal of Management Information Systems, 24(3), 45-78.
Shirley, S., & Hevner, A. (2013). POSITIONING AND PRESENTING DESIGN SCIENCE Types of Knowledge in Design Science Research. MIS Quarterly, 37(2), 337-355.
Yin, J., Lampert, A., Cameron, M., Robinson, B., & Power, R. (2012). Using Social Media to Enhance Emergency Situation Awareness. IEEE Intelligent Systems, 52–59-52–59.