This project aims to develop a cloud-based computing framework that will systematically monitor, collect and integrate disaster and crisis-related data streams from diverse social network channels, and turn them into actionable intelligence in a knowledge-base for intelligent inferencing, crisis handling, and decision making. The key components of the proposed framework are three: First, the collection of multilingual and multimodal data (text, audios, videos, images, and location information or other sensor-data) and their conversion into a unified machine-understandable dynamic knowledge-base. Second, the application of Natural Language Processing (NLP), Data Mining, Machine Learning, and Artificial Intelligence techniques for automatic identification, prediction, and detection of any type of crisis or disaster using the unified knowledge-base from the previous stage. 3) Third, the development of an interactive dashboard visualization to facilitate crisis monitoring and management.
•Dynamic knowledge extraction from heterogeneous sources, and information from social media channels to facilitate disaster-related operations and activities. •Apply NLP, artificial intelligence, and machine learning algorithms for automatic extraction of key features and information from crisis and disaster-related posts from news channels and social media (Twitter) to identify disaster type, what kind of key information can be identified through text analysis (Location, Time, Any Human Injury or Not ....) •Demonstrate the effectiveness of automatically extracted knowledge and information by applying such knowledge in creating values for the management of disaster-related activities.
The research team will provide all the resources required to complete this project.
Crawl a data source (e.g., Twitter, Google News ….etc.) on real-time Analyze news for binary classification (Disaster/emergency related or not) If a disaster/emergency, what type of Disaster it is? Based on the Disaster type, what kind of key information can be identified through text analysis (Location, Time, Any Human Injury or Not ....)
Analyze news for binary classification (Disaster/emergency related or not) If a disaster/emergency, what type of Disaster it is? Based on the Disaster type, what kind of key information can be identified through text analysis (Location, Time, Any Human Injury or Not ....) More Advanced/Interesting challenges if they want to take it further How to track or follow up a disaster over time and update such information (update the key information for better emergency response, for example) Credibility Analysis (Fake or Real information/news) Sentiment Analysis (Based on a disaster event and responses by the authorities, people may have reacted with different level of satisfaction – analyze such public sentiment based on comments etc.
Depending on the volunteer’s background and time, they may address one or more of the above tasks and investigate them to a certain depth. Anything they produce maybe adopted in our project with further enhancement or integration.
After implementation, the solution would be deployed with humanitarian organizations such as the emergency response department in the ministry of civil defense.
An interface will be provided to the public, where the public notification will be posted by humanitarian organizations.
Project scope revision