This post is also available in: heעברית (Hebrew)

The vast majority of tweets associated with an emergency with a hashtag are wishes of support and discussion about the event. However, on a much smaller scale, people in the emergency itself use Twitter to request aid, share resources, offer services, and provide first-hand reports of the situation. 

Conventional public access to Twitter tweets are limited to small volumes within short periods of time. Twitter offers commercial Enterprise Level access to APIs that allow large scale search and filtering over all Twitter data. 

The US National Institute of Standards and Technology (NIST) requires access and use on an Enterprise-Level API to search Twitter during emergencies, such as floods, wildfires, earthquakes, other natural disasters, in order to identify as many tweets coming from the emergency as possible. 

NIST’s current practice uses public APIs with multiple queries, but it will be much more effective at collecting all the “critical” tweets during an emergency if enterprise APIs, which have much higher limits, will be used.

As part of the Public Safety Communications Research (PSCR) program, NIST began TREC Incident Streams project to support research into social media analytics during emergencies and natural disasters. 

The project identifies tweets sent during emergencies and annotates them for priority and according to an emergency response ontology. NIST then invites research teams from anywhere in the world to develop software systems to automatically filter and classify the tweets. NIST develops the infrastructure to measure the effectiveness of those systems, and a forum for sharing research results, which speeds the improvement of the technology. When the technology matures, such systems could drive analytic dashboards or decision-making processes in public safety organizations. 

Emergency events identified for collection will have the following properties: have significant duration, involve public safety response, occur in places where English is a primary or significant language, etc. Demonstrations, protests, labor strikes, riots and/or large public events without an emergency safety component (like the Super Bowl or a major outdoor concert) are examples of events the TREC Incident Streams track will not collect. 

NIST will obtain samples of tweets that people send during emergency events and use them to measure the effectiveness of technology designed to filter social media streams during those emergency events and route actionable information to public safety personnel, according to