• Home
  • About Us
  • Contact Us
  • Privacy Policy
  • Special Offers
Business Intelligence Info
  • Business Intelligence
    • BI News and Info
    • Big Data
    • Mobile and Cloud
    • Self-Service BI
  • CRM
    • CRM News and Info
    • InfusionSoft
    • Microsoft Dynamics CRM
    • NetSuite
    • OnContact
    • Salesforce
    • Workbooks
  • Data Mining
    • Pentaho
    • Sisense
    • Tableau
    • TIBCO Spotfire
  • Data Warehousing
    • DWH News and Info
    • IBM DB2
    • Microsoft SQL Server
    • Oracle
    • Teradata
  • Predictive Analytics
    • FICO
    • KNIME
    • Mathematica
    • Matlab
    • Minitab
    • RapidMiner
    • Revolution
    • SAP
    • SAS/SPSS
  • Humor

AI uses tweets to help researchers analyze floods

April 26, 2019   Big Data

Social media gets a lot of negative press, but there’s more to Twitter and Facebook than botnets, memes, and political trolls. In a research paper preprint on Arxiv.org (“Integrating Social Media into a Pan-European Flood Awareness System: A Multilingual Approach“), scientists at the Joint Research Center, the European Commission’s science and knowledge service, detail a prototype — Social Media for Flood Risk (SMFR) — that “enriches” Europe’s Flood Awareness System (EFAS) with real-time reports from Twitter users.

It builds on research published by Harvard and Google in August 2018, which described an AI model capable of predicting the location of aftershocks up to one year after a major earthquake, and by Facebook AI researchers in December, who developed a method to analyze satellite imagery and to quantify damage from fires and other disasters. More recently, scientists at Google published a retrospective on a machine learning system that accurately predicts riverine floods — that is, floods from overrun riverbanks — with 75% precision.

Separately, researchers in the U.K. have used tweet-ingesting machine learning algorithms to map out where violence is likely to occur during riots, to project when mass protests might be imminent, and to identify gang members.

“Over the past decade, social media has emerged as a relevant information source during disasters, prompting researchers from diverse areas to converge on this domain,” the paper’s coauthors wrote. “Social media analysis has demonstrated the potential to provide timely, precious information about the spatial and temporal development of a crisis, as well as supporting the identification of key disaster-related events.”

 AI uses tweets to help researchers analyze floods

Above: Mapping tweets to flooding with AI.

First, a quick primer on the EFAS: It’s a part of the Copernicus Emergency Management Service (Copernicus EMS) and operated by the European Commission’s Emergency Response Coordination Centre (ERCC), a division of the European Commission’s Civil Protection and Humanitarian Aid Operations set up to support coordinated responses to disasters inside and outside Europe. Much like the U.S.’s Federal Emergency Management Agency, ERCC monitors hazards and risks, collects and analyzes data on disasters, and prepares plans for teams and equipment deployment. And ERCC sources EFAS for forecasts — principally probabilistic medium-range flood forecasts (including short-range flash floods), but also seasonal forecasts, impact assessments, and early warnings.

The researchers’ system tapped EFAS to determine when the risk of floods in a certain geographic area exceeded a threshold. This triggered data collection from social media — Twitter — to the tune of up to 400 keywords at a time, the public streamer API’s maximum limit.

Extracting messages with relevant keywords (i.e., words indicating a flood is about to happen or recently happened) was no easy task, given that EFAS covers an area where the population speaks more than 27 languages. The team’s solution was a multi-lingual classification system that used language-agnostic mathematical representation of words, or word embeddings, to infer similarities among keywords in four tongues: German, English, Spanish, and French.

To train it, they sourced a corpus containing over 7,000 annotated messages (between 1,200 and 2,300 messages per language). Meanwhile, they used a separate model to suss out “representative” messages (tweets having at least a 90% probability of being flood-related) for areas in which flood risk had been predicted.

To test the robustness of their approach, the scientists integrated SMFR into EFAS and deployed it during recent floods affecting Calabria, Italy, in early October 2018. SMFR collected two days’ worth of tweets — about 14,347 in all — which SMFR analyzed for “relatedness.” The researchers report that the AI-filtered messages closely correlated with actual flooding, and they say it’s a promising first step toward a system which could shorten response time in early stages of disasters.

“[D]uring the development of an event, collected messages could be valuable to international rescue coordinators … because they provide insights about the local response, about whether alerts that have been issued by authorities, and about some of the concerns that those affected by a flood or a flood alert may have,” the team wrote. “As future research activities, one can envision a global system comprising dozens of languages [and] further steps in the direction of using social media as a data source that can feed into a predictive model.”

Let’s block ads! (Why?)

Big Data – VentureBeat

analyze, floods, help, researchers, tweets, Uses
  • Recent Posts

    • We were upgraded to the Unified Interface for Dynamics 365. Now What?
    • Recreating Art – the unexpected way
    • Upcoming Webinar: Using Dynamics 365 to Empower your Marketing and Sales Teams with Digital Automation
    • Center for Applied Data Ethics suggests treating AI like a bureaucracy
    • Improving Dynamics 365 Data Integrations with Alternate Keys
  • Categories

  • Archives

    • January 2021
    • December 2020
    • November 2020
    • October 2020
    • September 2020
    • August 2020
    • July 2020
    • June 2020
    • May 2020
    • April 2020
    • March 2020
    • February 2020
    • January 2020
    • December 2019
    • November 2019
    • October 2019
    • September 2019
    • August 2019
    • July 2019
    • June 2019
    • May 2019
    • April 2019
    • March 2019
    • February 2019
    • January 2019
    • December 2018
    • November 2018
    • October 2018
    • September 2018
    • August 2018
    • July 2018
    • June 2018
    • May 2018
    • April 2018
    • March 2018
    • February 2018
    • January 2018
    • December 2017
    • November 2017
    • October 2017
    • September 2017
    • August 2017
    • July 2017
    • June 2017
    • May 2017
    • April 2017
    • March 2017
    • February 2017
    • January 2017
    • December 2016
    • November 2016
    • October 2016
    • September 2016
    • August 2016
    • July 2016
    • June 2016
    • May 2016
    • April 2016
    • March 2016
    • February 2016
    • January 2016
    • December 2015
    • November 2015
    • October 2015
    • September 2015
    • August 2015
    • July 2015
    • June 2015
    • May 2015
    • April 2015
    • March 2015
    • February 2015
    • January 2015
    • December 2014
    • November 2014
© 2021 Business Intelligence Info
Power BI Training | G Com Solutions Limited