Home BlogWeekly News 10 Bits: the Data News Hotlist

10 Bits: the Data News Hotlist

by Cassidy Chansirik

This week’s list of data news highlights covers February 27, 2021 – March 5, 2021 and includes articles about revealing disparities in traffic stops and modeling the effectiveness of double-masking. 

1. Using Data to Eliminate World Hunger

The World Food Program (WFP), the food assistance branch of the United Nations, is using geospatial data to determine where communities in need of support are located, what the best way to reach them is, and whether they are affected by natural disasters. WFP is using this map, together with existing data on food insecurity for different regions, to better target its interventions. 

2. Improving the Performance of Football Players

Football clubs in the English Premier League have turned to data analytics to improve the performance of their players. During training, teams are using wearable devices to measure players’ workload and fatigue to ensure external factors, such as amount of sleep, are not impacting their physical performance. Another tool uses data on passes, shots, and turnovers in combination with optical tracking, which pinpoints the position of players in relation to the ball or other players, to improve a team’s passing accuracy. 

3. Fighting Wildfires with Better Data

Cornea, a data service for disaster planning, response, and recovery, is combining geographical, weather, and historical data to guide firefighters on when to push forward or retreat during a wildfire. The organization is developing two tools: the Suppression Difficulty Index, which uses geographical locations to map and alert firefighters of wind or water conditions that could worsen a fire, and the Potential Control Lines tool, which uses typography and vegetation data to locate areas of a wildfire that are likely to be pushed back to reduce overall spread. 

4. Understanding How Social Media Can Influence Beliefs about COVID-19

Researchers at Northwestern University in Chicago have used machine learning to study how different events affect people’s beliefs about COVID-19. The team trained a model to categorize approximately 93 million tweets from 9 million users about COVID-19 based on their perceived susceptibility, severity, benefits, and barriers toward the virus. The team found that public attitudes toward the virus fluctuated equally around scientific events such as journal publications and non-scientific events, like speeches from politicians. 

5. Modeling the Effectiveness of Double Masking to Prevent the Spread of COVID-19

Researchers at Riken’s Center for Computational Science and Kobe University in Japan have used Fugaku, the world’s fastest supercomputer, to model the flow of virus particles between people wearing different types and numbers of masks. They found that a single surgical mask made of non-woven material effectively blocked particles 85 percent of the time, and adding a polyethylene fabric mask on top only increased effectiveness to 89 percent. 

6. Creating a New Self-Supervised Computer Vision AI Model

Researchers at Facebook have developed SEER, a computer vision model that learns from the information it is given without relying on labeled images from a dataset. Unlike previous models that were trained on labeled images, researchers trained SEER to identify objects from one billion unlabeled publicly-sourced Instagram photos. By reducing the time spent manually labeling images for training, the team hopes the technology can be applied to images on Facebook and Instagram feeds to automatically generate image descriptions, helping the company to better identify community policy violations. 

7. Emulating How Dogs Sniff Out Diseases Through Neural Networks

Researchers at the Massachusetts Institute of Technology have developed neural networks that emulate how dogs sniff out diseases, like prostate cancer, from urine samples. The team analyzed urine samples to identify the unique mixture of chemicals that produces the scent signatures indicative of prostate cancer, and used this information to train neural networks to identify prostate cancer using the molecular-level breakdown of the scent-producing chemicals. The networks identified positive cases with 71 percent accuracy and negative cases with up to 76 percent accuracy. 

8. Revealing the Contents of Historical Letters

Researchers at the Massachusetts Institute of Technology are using an algorithm to work out how best to unfold 600 unopened and undelivered letters sent from various European countries to the Hague in the Netherlands between 1680 and 1706. The team first scanned the folded letters with X-ray imaging to create a 3D reconstruction. Then they used an algorithm to identify individual layers of paper within each reconstruction and analyzed the thickness of paper creases to determine how to digitally unfold the letters. With this technique, the team read a letter from the French legal professional Jacques Sennacques, who requested an official death certificate for a relative. Until now, historians only knew the name of the recipient and not the content of the letter. 

9. Piloting a COVID-19 Vaccine Passport

The state of New York has collaborated with IBM to pilot a COVID-19 vaccine passport, which is an app that contains a unique QR code that businesses can scan to confirm a user’s vaccination status. Users can also choose what other types of medical information are stored within the app, such as COVID-19 tests results. This past week, the passport was piloted at the basketball game between the Brooklyn Nets and the Dallas Mavericks in order to inform government officials on how the passport will function once restrictions are rolled back on large events and large gatherings. 

10. Revealing Disparities in Traffic Stops in North Carolina

The North Carolina Criminal Justice Analysis Center has analyzed ten years of data on traffic stops that took place in the state between 2009 to 2019. It found that law enforcement officers searched Black drivers and their vehicles nearly twice as often as white drivers. The data also revealed that despite an overall decline in the number of traffic stops, the number of searches that took place during the stops increased from 31,865 in 2016 to 38,000 in 2019. Furthermore,  the rate of consent searches, which is when drivers and passengers agree to have their belongings inspected, declined from 50 percent to 22 percent in 2019. 

Image credit: Marcus Winkler


You may also like

Show Buttons
Hide Buttons