This week’s list of data news highlights covers September 3-9, 2016 and includes articles about how autonomous sailboats are collecting ocean data in the Bering Sea and a new technique for limiting the influence of ISIS propaganda on YouTube.
A team of researchers from the University of Michigan has been using open data and analytics to help solve the contaminated water crisis in Flint, Michigan. The researchers aggregated open data about land and water use, water testing, housing data, and other information relevant to Flint’s water crisis to develop statistical models that can predict which homes are at risk for contaminated water and prioritize efforts to replace water service lines causing the contamination. This predictive approach is valuable because the city’s records of water service lines, which could easily reveal which pipes have lead or contribute to other types of contamination, are incomplete or inconsistent for over 30 percent of homes.
The United States Pharmacopeial Convention (USP), a nonprofit standards-setting organization for food and pharmaceutical ingredients, has launched a new version of its Food Fraud Database called FFD 2.0 that allows users to easily track and interpret data about fraudulent adulteration of food products. FFD 2.0 tracks incident reports, regulatory and judicial records, scientific research, and other data from around the world to monitor incidents of fraud, such as when a company misleadingly substitutes ingredients with cheaper and potentially dangerous alternatives. USP created FFD 2.0 to help industry comply with food safety requirements, inform consumers, and reduce the prevalence of food fraud, which USP estimates costs the industry up to $15 billion per year around the world.
Saildrone, a company that builds autonomous sailboats, has two of its boats combing the Bering Sea equipped with a variety of scientific sensors to gather oceanographic data. The Saildrone boats use specialized sonar to count pollock populations (a type of fish) and track seals, which feed on the pollock, as well as collect other environmental data for researchers. Autonomous data-collecting boats like Saildrone’s can help researchers and climatologists gather large amounts of data in remote locations and in difficult conditions for a variety of valuable purposes, such as studying climate change, improving weather forecasts, and combating illegal fishing.
Mobile health device manufacturer AliveCor has launched a system that can help users manage their risk of having a stroke with data from wearable connected devices. AliveCor’s Kardia Mobile app for smartphones can analyze data from AliveCor’s handheld Kardia electrocardiogram (ECG), which measures a heart’s electrical activity, and a Bluetooth-connected blood pressure monitor to warn users if it detects an atrial fibrillation, which can substantially increase someone’s risk of stroke. AliveCor has also developed an application program interface (API) to allow users to export their data to electronic health record systems.
The United States Tennis Association (USTA) is using IBM’s Watson cognitive computing platform to power several different services during the U.S. Open tennis championships. The U.S. Open app uses Watson’s natural language processing ability to respond to users’ questions and help them navigate the tournament. USTA is also using Watson’s speech-to-text ability to quickly and automatically generate subtitles for videos of the tournament. And USTA photographers are using Watson’s visual recognition services to rapidly identify the tennis players and celebrities in their pictures to speed up the process of sorting through their photos.
Google DeepMind, Alphabet’s United Kingdom-based AI research division, has developed an AI program called WaveNet capable of producing human-sounding speech 50 percent more effectively than any existing system. DeepMind researchers trained WaveNet’s artificial neural network to learn how to produce individual sound waves in the same way a human voice does by having it analyze audio signals 16,000 times per second. In a test of WaveNet producing English and Mandarin, humans could identify that WaveNet’s voice was still artificial but substantially more natural sounding than other text-to-speech software.
Several states with prescription drug monitoring databases, which keep track of every time doctors prescribe addictive medications such as opioids, are sending doctors reports that compare their opioid prescription rates to other doctors’ so they can avoid over prescribing the drugs. After a two-year pilot in Arizona, for example, counties that sent out report cards to doctors saw a 10-percent drop in opiate prescriptions and a 4-percent drop in overdose deaths.
Spanish startup Narrativa has developed an AI system for generating news coverage focusing on minor-league sports, which rarely get more than minimal, local coverage. Narrativa’s algorithms analyze basic game data, such as points scored, and contextual information, such as a team’s winning streak, to produce natural-sound summaries of games. Typically minor-league sports do not receive much news coverage and when they do, it’s only in the local language, because reaching large audiences can be expensive and time consuming. By automating this process, Narrativa hopes to make it more cost effective to produce easily translatable sports coverage that can reach more fans.
Airbnb has announced it will take several data-driven steps to prevent discrimination on its platform after research found that hosts regularly discriminate against minority users by not renting them rooms. Beginning in 2017, if a host declines a renter by saying their home is unavailable on a particular date, Airbnb’s platform will prevent the host from renting to anyone else on that date to discourage hosts from rejecting renters based on potentially biased reasons. Similarly, Airbnb will encourage the use of instant bookings, which allows renters to automatically book a listing as long as the desired date is available, without input from the host. Finally, Airbnb will experiment with ways to reduce the visibility of renter’s profile pictures when hosts consider potential renters to reduce the opportunity for hosts to reject renters for biased reasons.
Google’s Jigsaw, a technology incubator and think tank, has developed a system called the Redirect Method that takes advantage of Google’s advertising algorithms to dissuade YouTube users from accessing propaganda videos the Islamic State (ISIS) uses to gain recruits. When a person uses Google to searched for keywords of phrases Jigsaw has identified people supportive of ISIS often use, the Redirect Method will populate the search results with YouTube channels in Arabic and English with content that undermines ISIS propaganda. In a two-month pilot, the Redirect Method drew over 300,000 people to the the anti-ISIS videos and those users watched some of the channels twice as long as the average user’s YouTube viewing time. Jigsaw will launch another pilot this month to gauge the Redirect Method’s effectiveness for countering both ISIS and violent white supremacist propaganda in North America.
Image: Edwin Martinez.