|
Rego Granlund, & Helena Granlund. (2011). GPS impact on performance, response time and communication – A review of three studies. In E. Portela L. S. M.A. Santos (Ed.), 8th International Conference on Information Systems for Crisis Response and Management: From Early-Warning Systems to Preparedness and Training, ISCRAM 2011. Lisbon: Information Systems for Crisis Response and Management, ISCRAM.
Abstract: This paper describes the basic work performance analysis from three research projects with a goal to investigate the impact of a decision support system that presents global positioning system (GPS) information to the decision makers in crisis management organizations. The goal was to compare the performance between teams that had access to GPS information in the command post with teams that had access only to paper maps. The method used was controlled experiments with the C3Fire micro-world. A total of 304 participants, forming 48 teams, participated in the three studies. The participants came from three different groups, university students, municipal crisis management organizations and rescue service personnel. The result shows that the performance and communication change depending on if the teams used GPS support or paper maps. The result also shows that the participants' background and perceived complexity of the task have an impact on the results.
|
|
|
Tim J. Grant. (2005). Integrating sensemaking and response using planning operator induction. In B. C. B. Van de Walle (Ed.), Proceedings of ISCRAM 2005 – 2nd International Conference on Information Systems for Crisis Response and Management (pp. 89–96). Brussels: Royal Flemish Academy of Belgium.
Abstract: Emergency response managers often try to prepare plans in advance of foreseeable crises. However, pre-prepared plans almost never fit the precise situation when an actual crisis arises. This paper focuses on what can be done once the plan has been found to be a mismatch with the actual situation, and particularly on sensemaking and its integration into generating a response to a crisis situation. This paper proposes planning operator induction as a suitable technique. The Air Traffic Control and military air defence situation over the continental US on September 11, 2001, serves as the illustrative application. Since the paper reports research in progress, it closes by outlining the additional functionality that will have to be added to the existing POI implementation to produce a sensemaking tool.
|
|
|
Tim J. Grant, & Ariën J. Van Der Wal. (2012). A taxonomy of market mechanisms for information sharing in crisis response coalitions. In Z.Franco J. R. L. Rothkrantz (Ed.), ISCRAM 2012 Conference Proceedings – 9th International Conference on Information Systems for Crisis Response and Management. Vancouver, BC: Simon Fraser University.
Abstract: Information sharing is crucial in responding to and managing crises. A great variety of organizations respond to crises, ranging from international organizations, through non-governmental organizations, commercial suppliers, government ministries, and the media, to individual volunteers. At best, they collaborate to combine their respective strengths in creating synergy. At worst, they deconflict their operations so that the one does not get in the way of the other. In either case, the organizations must communicate with one another. Previous research has shown that this communication takes the form of an information market. This paper considers the possible mechanisms for information markets in humanitarian and military coalitions. While there has been work on market mechanisms in the economics literature, this needs to be adapted for the exchange of information and information services instead of physical goods and services. This paper proposes a taxonomy of market mechanisms that is tailored to information sharing within coalitions. © 2012 ISCRAM.
|
|
|
Tim J. Grant, F.L.E. Geugies, & Peter A. Jongejan. (2013). Social media in command & control: A proof-of principle experiment. In J. Geldermann and T. Müller S. Fortier F. F. T. Comes (Ed.), ISCRAM 2013 Conference Proceedings – 10th International Conference on Information Systems for Crisis Response and Management (pp. 52–61). KIT; Baden-Baden: Karlsruher Institut fur Technologie.
Abstract: The literature on the organizational use of social media in crisis response and management is largely concerned with communication between organizations and the general public (“citizens”). By contrast, there are few papers on the use of social media within organizations for operational purposes. One essential operational capability in such organizations is Command & Control (C2) or its equivalent. Our research focuses on the use of social media in C2 for crisis management at the operational and tactical levels. To enable the use of social media in C2, Jongejan and Grant (2012) extended Reuter, Marx and Pipek's (2011) theoretical framework. In the research reported in this paper, the extended framework was tested by performing a proof-of-principle experiment for a famine relief scenario with human subjects in the laboratory. The results show that more extensive concept demonstrations and field experimentation are justified.
|
|
|
Grégoire Burel, Lara S. G. Piccolo, Kenny Meesters, & Harith Alani. (2017). DoRES -- A Three-tier Ontology for Modelling Crises in the Digital Age. In eds Aurélie Montarnal Matthieu Lauras Chihab Hanachi F. B. Tina Comes (Ed.), Proceedings of the 14th International Conference on Information Systems for Crisis Response And Management (pp. 834–845). Albi, France: Iscram.
Abstract: During emergency crises it is imperative to collect, organise, analyse and share critical information between individuals and humanitarian organisations. Although dierent models and platforms have been created for helping these particular issues, existing work tend to focus on only one or two of the previous matters. We propose the DoRES ontology for representing information sources, consolidating it into reports and then, representing event situation based on reports. Our approach is guided by the analysis of 1) the structure of a widely used situation awareness platform; 2) stakeholder interviews, and; 3) the structure of existing crisis datasets. Based on this, we extract 102 dierent competency questions that are then used for specifying and implementing the new three-tiers crisis model. We show that the model can successfully be used for mapping the 102 dierent competency questions to the classes, properties and relations of the implemented ontology.
|
|
|
Anna Gryszkiewicz, & Fang Chen. (2010). Design requirements for information sharing in a crisis management command and control centre. In C. Zobel B. T. S. French (Ed.), ISCRAM 2010 – 7th International Conference on Information Systems for Crisis Response and Management: Defining Crisis Management 3.0, Proceedings. Seattle, WA: Information Systems for Crisis Response and Management, ISCRAM.
Abstract: Good support for information sharing and processing is essential for successful crisis management. A crisis manager handles information from many different sources and collaborates with many different actors. This study is therefore focusing on specifying some needs and requirements for information support systems for crisis management command and control centres. The study is based on case studies and interviews in Sweden with the aim to understand how information processing best can be supported from a crisis manager's perspective. The needs and requirements found in the study can be used in future system design or improvement.
|
|
|
Guido Te Brake, Tjerk De Greef, Jasper Lindenberg, Jouke Rypkema, & Nanja Smets. (2006). Developing adaptive user interfaces using a game-based simulation environment. In M. T. B. Van de Walle (Ed.), Proceedings of ISCRAM 2006 – 3rd International Conference on Information Systems for Crisis Response and Management (pp. 6–10). Newark, NJ: Royal Flemish Academy of Belgium.
Abstract: In dynamic settings, user interfaces can provide more optimal support if they adapt to the context of use. Providing adaptive user interfaces to first responders may therefore be fruitful. A cognitive engineering method that incorporates development iterations in both a simulated and a real-world environment is used to develop new adaptive concepts. In a simulated 3D-world, created with the Unreal Tournament game-engine, a team of emergency personnel have to rescue people and develop an understanding of the situation. We believe a game-based simulation environment can provide an effective platform for experiments in which crisis management situations can be created under controlled circumstances. Using this simulation, support concepts based on adaptive user interfaces can be developed and evaluated before they are implemented in a real-world setting. This paper describes the work that has been done, and presents the design of the planned experiments.
|
|
|
Shubham Gupta, & Craig A. Knoblock. (2010). Building geospatial mashups to visualize information for crisis management. In C. Zobel B. T. S. French (Ed.), ISCRAM 2010 – 7th International Conference on Information Systems for Crisis Response and Management: Defining Crisis Management 3.0, Proceedings. Seattle, WA: Information Systems for Crisis Response and Management, ISCRAM.
Abstract: In time-sensitive environments such as disaster management, decision-making often requires rapidly gathering the information from diverse data sources and then visualizing the collected information to understand it. Thus, it is critical to reduce the overhead in data integration and visualization for efficient decision-making. Geospatial mashups can be an effective solution in such environments by providing an integrated approach to extract, integrate and view diverse information. Currently, mashup building tools exist for creating mashups, but none of them deal with the issue of data visualization. An improper visualization of the data could result in users wasting precious time to understand the data. In this paper, we introduce a programming-by-demonstration approach to data visualization in geospatial mashups that allows the users to customize the data visualization.
|
|
|
Olof Görnerup, Per Kreuger, & Daniel Gillblad. (2013). Autonomous accident monitoring using cellular network data. In J. Geldermann and T. Müller S. Fortier F. F. T. Comes (Ed.), ISCRAM 2013 Conference Proceedings – 10th International Conference on Information Systems for Crisis Response and Management (pp. 638–646). KIT; Baden-Baden: Karlsruher Institut fur Technologie.
Abstract: Mobile communication networks constitute large-scale sensor networks that generate huge amounts of data that can be refined into collective mobility patterns. In this paper we propose a method for using these patterns to autonomously monitor and detect accidents and other critical events. The approach is to identify a measure that is approximately time-invariant on short time-scales under regular conditions, estimate the short and long-term dynamics of this measure using Bayesian inference, and identify sudden shifts in mobility patterns by monitoring the divergence between the short and long-term estimates. By estimating long-term dynamics, the method is also able to adapt to long-term trends in data. As a proof-of-concept, we apply this approach in a vehicular traffic scenario, where we demonstrate that the method can detect traffic accidents and distinguish these from regular events, such as traffic congestions.
|
|
|
Daniel Hahn. (2007). Non-restrictive linking in wireless sensor networks for industrial risk management. In K. Nieuwenhuis P. B. B. Van de Walle (Ed.), Intelligent Human Computer Systems for Crisis Response and Management, ISCRAM 2007 Academic Proceedings Papers (pp. 605–609). Delft: Information Systems for Crisis Response and Management, ISCRAM.
Abstract: The OSIRIS project addresses the disaster management workflow in the phases of risk monitoring and crisis management. Risk monitoring allows the continuous observation of endangered areas combined with sensor deployment strategies. The crisis management focuses on particular events and the support by sensor networks. Four complementary live demonstrations will validate the OSIRIS approach. These demonstrations include water contamination, air pollution, south European forest fire, and industrial risk monitoring. This paper focuses on the latter scenario: the industrial risk monitoring. This scenario offers the special opportunity to demonstrate the relevance of OSIRIS by covering all the aspects of monitoring, preparation and response phases of both environmental risk and crisis management. The approach focuses on non-restrictive linking in a wireless sensor network in order to facilitate the addition and removal of nodes providing open interaction primitives allowing the comfortable integration, exclusion, and modification. A management layer with an event-triggered and service-based middleware is proposed. A live lab with real fire is illustrated.
|
|
|
Dan Harnesk, Sören Samulesson, & John Lindström. (2009). Action research supported implementation of a crisis competence centre. In S. J. J. Landgren (Ed.), ISCRAM 2009 – 6th International Conference on Information Systems for Crisis Response and Management: Boundary Spanning Initiatives and New Perspectives. Gothenburg: Information Systems for Crisis Response and Management, ISCRAM.
Abstract: The purpose of this paper is to describe the development of a regional crisis competence centre in Sweden. The overall research approach utilizes action research methodology to support the development process. The overall development process is divided into three steps; early need finding, the organizational platform, and business development. This paper is devoted to the first phase – early need finding – which deals with the specific needs and demand of four municipality organizations and how these requirements can be converged into a organizational setting that correspond to national crisis management strategies as well as regional responsibilities in crisis management. The research approach is action based and the theoretical considerations to support progress in development activities are from the Actor Network Theory (ANT). The main conclusion from the research is that the formation of a crisis competence centre revolves around sharing competences, and compromising between internal and external objectives in the progress towards a common goal.
|
|
|
Janine Hellriegel, & Michael Klafft. (2014). A tool for the simulation of alert message propagation in the general population. In and P.C. Shih. L. Plotnick M. S. P. S.R. Hiltz (Ed.), ISCRAM 2014 Conference Proceedings – 11th International Conference on Information Systems for Crisis Response and Management (pp. 65–69). University Park, PA: The Pennsylvania State University.
Abstract: Informing and alerting the population in disaster situations is a challenging task. Numerous situational factors have to be considered, as well as the impact of a plethora of communication channels, and multiplication effects in the population. In order to optimize the alerting strategies and enhance alert planning, it would be beneficial to model the dissemination of alerts. In this paper, we present a general overview of an alert dissemination model as well as its prototypical implementation in a simulation software. The software takes situational parameters such as time of day and location into account and can even infer characteristics of the alerting infrastructure from geospatial information.
|
|
|
Hemant Purohit, & Jennifer Chan. (2017). Classifying User Types on Social Media to inform Who-What-Where Coordination during Crisis Response. In eds Aurélie Montarnal Matthieu Lauras Chihab Hanachi F. B. Tina Comes (Ed.), Proceedings of the 14th International Conference on Information Systems for Crisis Response And Management (pp. 656–665). Albi, France: Iscram.
Abstract: Timely information is essential for better dynamic situational awareness, which leads to efficient resource planning, coordination, and action. However, given the scale and outreach of social media�a key information sharing platform during crises, diverse types of users participate in discussions during crises, which affect the vetting of information for dynamic situational awareness and response coordination activities. In this paper, we present a user analysis on Twitter during crises for three major user types�Organization, Organizationaffiliated (a person�s self-identifying affiliation with an organization in his/her profile), and Non-affiliated (person not identifying any affiliation), by first classifying users and then presenting their communication patterns during two recent crises. Our analysis shows distinctive patterns of the three user types for participation and communication on social media during crises. Such a user-centric approach to study information sharing during crisis events can act as a precursor to deeper domain-driven content analysis for response agencies.
|
|
|
Hemant Purohit, & Kathleen Moore. (2018). The Digital Crow's Nest: A Framework for Proactive Disaster Informatics & Resilience by Open Source Intelligence. In Kees Boersma, & Brian Tomaszeski (Eds.), ISCRAM 2018 Conference Proceedings – 15th International Conference on Information Systems for Crisis Response and Management (pp. 949–958). Rochester, NY (USA): Rochester Institute of Technology.
Abstract: The research on technology-assisted crisis management has primarily existed for two decades since 9/11. Although, the focus of technology research has been centered around tools to assist the response phase after a disaster. There has been a lack of emphasis on the role and design of technologies to assist the other phases of the crisis management cycle, particularly preparedness and mitigation phases to lead towards the vision of building resilient communities. In this paper, we first identify resilience characteristics of a community from the prior literature. We then analyze a co-occurrence network of concepts in the ISCRAM publications to validate the gap in relating technologies to resilience and conduct an indicator analysis of factors for proactive disaster informatics via a case study of recent disaster. Our analysis leads us to propose a conceptual framework “Digital Crow's Nest” based on Open Source Intelligence to improve the technology design for community resilience.
|
|
|
Hendrik Stange, Sylvia Steenhoek, Sebastian Bothe, & François Schnitzler. (2015). Insight-driven Crisis Information ? Preparing for the Unexpected using Big Data. In L. Palen, M. Buscher, T. Comes, & A. Hughes (Eds.), ISCRAM 2015 Conference Proceedings ? 12th International Conference on Information Systems for Crisis Response and Management. Kristiansand, Norway: University of Agder (UiA).
Abstract: National information and situation centers are faced with rising information needs and the question of how to prepare for unexpected situations. One promising development is the access to vastly growing data produced by distributed sensors and a socially networked society. Current emergency information systems are limited in the amount of complex data they can process and interpret in real-time and provide only partially integrated prediction and alarming capabilities. In this paper we present a novel approach to build a new type of automated and scalable information systems that intelligently make use of massive streams of structured and unstructured data and incorporate human feedback for automated incident detection and learning. Big data technologies, uncertainty handling and privacy-by-design are employed to match end-user system requirements. We share first experiences analyzing data from the centennial flood in Germany 2013.
|
|
|
Benjamin Herfort, João Porto De Albuquerque, Svend-Jonas Schelhorn, & Alexander Zipf. (2014). Does the spatiotemporal distribution of tweets match the spatiotemporal distribution of flood phenomena? A study about the River Elbe Flood in June 2013. In and P.C. Shih. L. Plotnick M. S. P. S.R. Hiltz (Ed.), ISCRAM 2014 Conference Proceedings – 11th International Conference on Information Systems for Crisis Response and Management (pp. 747–751). University Park, PA: The Pennsylvania State University.
Abstract: In this paper we present a new approach to enhance information extraction from social media that relies upon the geographical relations between twitter data and flood phenomena. We use specific geographical features like hydrological data and digital elevation models to analyze the spatiotemporal distribution of georeferenced twitter messages. This approach is applied to examine the River Elbe Flood in Germany in June 2013. Although recent research has shown that social media platforms like Twitter can be complementary information sources for achieving situation awareness, previous work is mostly concentrated on the classification and analysis of tweets without resorting to existing data related to the disaster, e.g. catchment borders or sensor data about river levels. Our results show that our approach based on geographical relations can help to manage the high volume and velocity of social media messages and thus can be valuable for both crisis response and preventive flood monitoring.
|
|
|
Herkevall, J., & Johansson Björn J E. (2023). Is this a date? – New perspectives on ICT for harmonized inter-organizational crisis management. In Jaziar Radianti, Ioannis Dokas, Nicolas Lalone, & Deepak Khazanchi (Eds.), Proceedings of the 20th International ISCRAM Conference (pp. 67–75). Omaha, USA: University of Nebraska at Omaha.
Abstract: In this paper, we challenge common conceptions related to the role and development of information and communication technology (ICT) for crisis management. Based on an understanding of inter-organizational crisis management as self-organizing through processes of harmonization in complex adaptive systems, ICT is positioned as an enabler in the bridging of social and organizational boundaries. In this view, the primary requirements for ICT in inter-organizational settings are defined by current information sharing needs, which are continuously changing with the context and co-working partners. From this understanding of inter-organizational relations in crisis management, this paper suggests two complementary approaches to ICT focusing on policy adjustment and training that supports adaptive organizational capabilities and utilization of easily available commercial ICT. The ideas put forth in this paper are intended to add perspectives and spark discussions on ICT for crisis management.
|
|
|
Josune Hernantes, Jose M. Torres, Ana Laugé, Jose Mari Sarriegi, Iztok Starc, Eva Zupancic, et al. (2010). Using GMB methodology on a large crisis model. In C. Zobel B. T. S. French (Ed.), ISCRAM 2010 – 7th International Conference on Information Systems for Crisis Response and Management: Defining Crisis Management 3.0, Proceedings. Seattle, WA: Information Systems for Crisis Response and Management, ISCRAM.
Abstract: Mitigating, detecting, evaluating, responding and recovering from crises are highly complex tasks that involve many decision makers (agents). As a consequence using collaborative methods that allow the cooperation among these agents during the crisis management strategy and procedures design is of significant importance. Group Model Building (GMB) is a robust collaborative methodology that has been successfully used for modelling several complex socio-technical problems, where different agents may have diverse perspectives or interests in the problem under analysis. Through the development of a series of exercises, GMB allows the integration of these initially fragmented perspectives. Modellers translate the knowledge elicited from experts during GMB workshops into simulation models that reproduce the behaviour of the problem. This paper presents the use and adaptation of the GMB methodology in a research project about large pan European crises due to outages in the electricity sector.
|
|
|
Herrera, L. C., & Gjøsæter, T. (2023). Leveraging Crisis Informatics Experts: A co-creating approach for validation of social media research insights. In Jaziar Radianti, Ioannis Dokas, Nicolas Lalone, & Deepak Khazanchi (Eds.), Proceedings of the 20th International ISCRAM Conference (pp. 439–448). Omaha, USA: University of Nebraska at Omaha.
Abstract: Validation of findings is a challenge in practice-based research. While analysis is being conducted and findings are being constructed out of data collected in a defined period, practitioners continue with their activities. This issue is exacerbated in the field of crisis management, where high volatility and personnel turnover make the capacity to attend research demands scarce. Therefore, conducting classic member validation is logistically challenging for the researcher. The need for rigor and validity calls for alternative mechanisms to fulfill requirements for academic research. This article presents an approach for validation of results of a qualitative study with public organizations that use social media as a source of information in the context of crisis management. The unavailability of original interview-objects to validate our findings resulted in an alternative validation method that leveraged experts in crisis informatics. By presenting our approach, we contribute to encouraging rigor in qualitative research while maintaining the relationship between practice and academia.
|
|
|
Thomas Heverin, & Lisl Zach. (2010). Microblogging for crisis communication: Examination of twitter use in response to a 2009 violent crisis in the Seattle-Tacoma, Washington area. In C. Zobel B. T. S. French (Ed.), ISCRAM 2010 – 7th International Conference on Information Systems for Crisis Response and Management: Defining Crisis Management 3.0, Proceedings. Seattle, WA: Information Systems for Crisis Response and Management, ISCRAM.
Abstract: This research-in-progress paper reports on the use of microblogging as a communication and information sharing resource during a recent violent crisis. The goal of the larger research effort is to investigate the role that microblogging plays in crisis communication during violent events. The shooting of four police officers and the subsequent 48-hour search for the suspect that took place in the Seattle-Tacoma area of Washington in late November 2009 is used as a case study. A stream of over 6,000 publically available messages on Twitter, a popular microblogging site, was collected and individual messages were categorized as information, opinion, technology, emotion, and action-related. The coding and statistical analyses of the messages suggest that citizens use microblogging as one method to organize and disseminate crisis-related information. Additional research is in progress to analyze the types of information transmitted, the sources of the information, and the temporal trends of information shared.
|
|
|
Michael Hiete, & Mirjam Merz. (2009). An indicator framework to assess the vulnerability of industrial sectors against indirect disaster losses. In S. J. J. Landgren (Ed.), ISCRAM 2009 – 6th International Conference on Information Systems for Crisis Response and Management: Boundary Spanning Initiatives and New Perspectives. Gothenburg: Information Systems for Crisis Response and Management, ISCRAM.
Abstract: Natural and man-made hazards may affect industrial production sites by both direct losses (due to physical damage to assets and buildings) and indirect losses (production losses). Indirect losses, e.g. from production downtimes, can exceed direct losses multiple times. Thus, the vulnerability of industrial sectors to indirect losses is an important component of risk and its determination is an important part within risk analysis. In this paper a conceptual indicator framework is presented which allows to assess the indirect vulnerability of industrial sectors to different types of disasters in a quantitative manner. The results are useful for information sharing and decision making in crisis management and emergency planning (mitigation measures, business continuity planning), since the developed indicator system helps to take the complex phenomenon of industrial vulnerability and the underlying interdependencies into account. Besides the identification and conceptual motivation of the indicators, methodical aspects such as standardization, weighting and aggregation are addressed.
|
|
|
Hoang Nam Ho, Mourad Rabah, Ronan Champagnat, & Frédéric Bretrand. (2019). Towards an Automatic Assistance in Crisis Resolution with Process Mining. In Z. Franco, J. J. González, & J. H. Canós (Eds.), Proceedings of the 16th International Conference on Information Systems for Crisis Response And Management. Valencia, Spain: Iscram.
Abstract: To deal with a crisis situation, experts must undertake a chain of activities, called process, to minimize crisis
consequences. To assist the expert in making decision in crisis resolutions, authors propose a method aiming at
discovering crisis response processes. This method is based on a two-step strategy: the first step classifies the
system?s traces, representing stakeholders? past actions, into different sets, where each one represents a set of
response processes according to a specific context; the second step uses process mining algorithm to discover
the corresponding response plan process model based on the obtained chain of activities for each previously
classified context. These response plans will be a referenced aid for experts while making crisis resolution,
according to each context. The proposed approach is illustrated on the traces issued from the crisis caused by the
2010 Xynthia storm in France.
|
|
|
Gesine Hofinger, Robert Zinke, & Laura Künzer. (2011). Psychological requirements for crisis and emergency decision-support systems for public transport control centers. In E. Portela L. S. M.A. Santos (Ed.), 8th International Conference on Information Systems for Crisis Response and Management: From Early-Warning Systems to Preparedness and Training, ISCRAM 2011. Lisbon: Information Systems for Crisis Response and Management, ISCRAM.
Abstract: Control center staff is used to working with software applications for e.g. surveillance of production processes, for controlling and timing of industrial logistics, and for recording and filing incidents and actions. Yet, decision-support systems for emergency situations pose additional demands on employees in this domain. This paper reports first findings for psychological requirements for decision support systems in the rise of emergencies as identified in a federal German research project in the domain of public transportation. In control centers both the humans on duty who have to intervene during an emergency, and the technical system which provides decision-alternatives for supporting the action are considered. Based on findings of the project, psychological, technical and organizational requirements identified in interviews, observations, document analysis and additional relevant literature are generalized.
|
|
|
Hongmin Li, Doina Caragea, & Cornelia Caragea. (2021). Combining Self-training with Deep Learning for Disaster Tweet Classification. In Anouck Adrot, Rob Grace, Kathleen Moore, & Christopher W. Zobel (Eds.), ISCRAM 2021 Conference Proceedings – 18th International Conference on Information Systems for Crisis Response and Management (pp. 719–730). Blacksburg, VA (USA): Virginia Tech.
Abstract: Significant progress has been made towards automated classification of disaster or crisis related tweets using machine learning approaches. Deep learning models, such as Convolutional Neural Networks (CNN), domain adaptation approaches based on self-training, and approaches based on pre-trained language models, such as BERT, have been proposed and used independently for disaster tweet classification. In this paper, we propose to combine self-training with CNN and BERT models, respectively, to improve the performance on the task of identifying crisis related tweets in a target disaster where labeled data is assumed to be unavailable, while unlabeled data is available. We evaluate the resulting self-training models on three crisis tweet collections and find that: 1) the pre-trained language model BERTweet is better than the standard BERT model, when fine-tuned for downstream crisis tweets classification; 2) self-training can help improve the performance of the CNN and BERTweet models for larger unlabeled target datasets, but not for smaller datasets.
|
|
|
Hongmin Li, Xukun Li, Doina Caragea, & Cornelia Caragea. (2018). Comparison of Word Embeddings and Sentence Encodings for Generalized Representations in Crisis Tweet Classifications. In Kristin Stock, & Deborah Bunker (Eds.), Proceedings of ISCRAM Asia Pacific 2018: Innovating for Resilience – 1st International Conference on Information Systems for Crisis Response and Management Asia Pacific. (pp. 480–493). Albany, Auckland, New Zealand: Massey Univeristy.
Abstract: Many machine learning and natural language processing techniques, including supervised and domain adaptation algorithms, have been proposed and studied in the context of filtering crisis tweets. However, applying these approaches in real-time is still challenging because of time-critical requirements of emergency response operations and also diversities and unique characteristics of emergency events. In this paper, we explore the idea of building “generalized” classifiers for filtering crisis tweets that can be pre-trained, and are thus ready to use in real-time, while generalizing well on future disasters/crises data. We propose to achieve this using simple feature based adaptation with tweet representations based on word embeddings and also sentence-level embeddings, representations which do not rely on unlabeled data to achieve domain adaptations and can be easily implemented. Given that there are different types of word/sentence embeddings that are widely used, we propose to compare them to get a general idea about which type works better with crisis tweets classification tasks. Our experimental results show that GloVe embeddings in general work better with the datasets used in our evaluation, and that the supervised algorithms used in our experiments benefit from GloVe embeddings trained specifically on crisis data. Furthermore, our experimental results show that following GloVe, the sentence embeddings have great potential in crisis tweet tasks.
|
|