Event-Related Bias Removal for Real-time Disaster Events

Findings of the Association for Computational Linguistics: EMNLP 2020, pages 3858–3868 November 16 - 20, 2020. c⃝2020 Association for Computational Linguistics 3858 Event-Related Bias Removal for Real-time Disaster Events Evangelia Spiliopoulou∗ Salvador Medina Maza∗ Eduard Hovy Alexander Hauptmann Language Technologies Institute Carnegie Mellon University {espiliop,salvadom,ehovy,alex}@cs.cmu.edu Abstract Social media has become an important tool to share information about crisis events such as natural disasters and mass attacks. Detect- ing actionable posts that contain useful infor- mation requires rapid analysis of huge vol- ume of data in real-time. This poses a com- plex problem due to the large amount of posts that do not contain any actionable information. Furthermore, the classification of information in real-time systems requires training on out- of-domain data, as we do not have any data from a new emerging crisis. Prior work fo- cuses on models pre-trained on similar event types. However, those models capture unnec- essary event-specific biases, like the location of the event, which affect the generalizability and performance of the classifiers on new un- seen data from an emerging new event. In our work, we train an adversarial neural model to remove latent event-specific biases and im- prove the performance on tweet importance classification. 1 Introduction Effective management of crisis situations like nat- ural disasters (e.g. earthquakes, floods) or attacks (e.g. bombings, shootings) is an extremely sensi- tive and complex phenomenon that requires effi- cient coordination of people from multiple disci- plines along with proper allocation of time and re- sources (Tapia et al., 2011a; Maitland et al., 2009). Given that we live in the era of information and social media, filtering important nuggets of infor- mation from real-time data and using them into decision-making constitutes a crucial research di- rection (Tapia et al., 2011b). Critical information from social media is found only in small amounts. Hence it is difficult to ex- * Equal contribution. tract and analyze the data stream, since it is impossi- ble to manually process the amount of information shared in social media in real-time. Therefore, it is important to detect data that contain useful informa- tion for decision-making and automatically extract it (Sutton et al., 2008; Palen et al., 2010). Even though sentence classification is a well-studied NLP problem, common approaches do not bring the expected results (Reuter et al., 2018). The main reason why common approaches fail is the lack of in-domain data (Mccreadie et al., 2019; Hiltz et al., 2014). Most emerging crisis are unex- pected and data analysis must be done real-time, within a small time-frame (Plotnick et al., 2015). Even if we might have high quality annotated data from previous similar crisis situations, we will not have data from the emerging event that we want to classify. For example, let us assume an earthquake in Seattle happens right now. Although we may have annotated data from a previous earthquake in Los Angeles, most of the parameters would be entirely different (e.g. location names, damages, times, etc) since the cities and populations differ. Furthermore, because some of …

Content truncated to 500 words