Difference between revisions of "DisasterRecord"
Revision as of 21:15, 5 November 2018
The DisasterRecord tool, our CallForCode submission, processes social media and crowdsourced data including imagery gathered during disaster events to support both individuals and community responders.
DisasterRecord substantially reduces the burden of analysis, interpretation, and decision making in comparison with tools like Ushahidi and Sahana Foundation Software. It enables automatic filtering, categorizing, and crucially, geolocating textual as well as image objects of social media data. Additionally, the tool analyzes geographical data and integrates satellite imagery processed for relevant features (such as flooded urban areas) for better decision making. All social media data processing is in real-time. Thus, DisasterRecord remains up-to-date to meet real-time coordination during the unfolding of a disaster. The iterative design of the tool, along with our ability to test and demonstrate it with real-world data (millions of tweets, background knowledge, satellite images, etc.) of disasters (such as Kerala flood 2018, Houston flood 2016, and Chennai flood 2015) allowed us to benefit from professional feedback of some of the most active humanitarian and non-profit organizations in the area, e.g., Joint IDP Profiling Service, Nepal Monitor, and Digital Humanitarian Network.
DisasterRecord meets the requirements of a variety of users. A humanitarian organization may analyze the situation at a community level for deploying and mobilizing necessary help. A first response coordinator can monitor a specific type of emergency needs. Affected individuals may need to know about the nearest available help. Persons wishing to provide support can identify current needs in the geographic proximity for the type of help they can provide.
In the aggregation level of analysis, DisasterRecord analyzes, categorizes, and geolocates location names extracted from non-georeferenced tweets. It visualizes aggregated information in a selected geographical bounding box. DisasterRecord eliminates the low-level clutter to surface essential information about disasters to responders. It also abstracts and categorizes data to provide community-level analysis and insights. Upon choosing a location of interest, the key aggregated information includes:
- Location-specific textual data:
- Categorizing tweets to need types
- Aggregating tweets with respect to location vicinity.
- Location-specific image features:
- Filtering for ‘Flooded images’
- Detecting objects of interest to assess severity and needs, including objects like people (to allow for learning about the disaster impact), animals (for monitoring livestock), and vehicles (to assess traffic maneuverability).
- Location-specific available help:
- Categorization of OpenStreetMap features to types and their counts.
- Location-specific thematic profile:
- Highlighting and summarizing the most discussed concepts in a region.
In the individual level analysis, DisasterRecord preserves low-level details to provide individuals with the situational awareness and the kind of available help for their needs. These include:
- Flooded areas around their location vicinity
- The possible available help around their location matching their needs
- Route guidance of unfolded routes for relief workers and individuals seeking help.
We used IBM Cloud Services in addition to our locally developed APIs and research components to support the functionalities of DisasterRecord.
Although we did not win the competition, we are thrilled to receive the following feedback from IBM- the competition sponsor:
“Great job on the project. It progressed very far in the competition and was definitely a great contribution"
Shruti Kar, Hussein S. Al-Olimat, Krishnaprasad Thirunarayan, Amit Sheth, Valerie Shalin, Dipesh Kadariya, and Mike Partin
- Shruti Kar, Hussein S. Al-Olimat, Krishnaprasad Thirunarayan, Valerie Shalin, Amit Sheth, and Srinivasan Parthasarathy. "D-record: Disaster Response and Relief Coordination Pipeline". In Proceedings of the ACM SIGSPATIAL International Workshop on Advances in Resilient and Intelligent Cities (ARIC 2018). ACM, 2018.
- Hussein S. Al-Olimat, Krishnaprasad Thirunarayan, Valerie Shalin, and Amit Sheth. Location Name Extraction from Targeted Text Streams using Gazetteer-based Statistical Language Models. The 27th International Conference on Computational Linguistics (COLING 2018).
- Jiongqian Liang, Peter Jacobs and Srinivasan Parthasarathy (2016). Human Guided Flood Mapping on Satellite Images. ACM SIGKDD Interactive Data Exploration and Analytics.