This week’s list of data news highlights covers May 28 – June 3, 2016 and includes articles about a system of wearable technologies that can help prevent asthma attacks and a new search engine for satellite imagery.
The Nasdaq stock exchange is testing an artificial intelligence system, originally designed to catch terrorists and sex traffickers, to analyze stock market activity and detect financial crimes such as insider trading and market manipulation. The system combines trading data with traders’ email and online chat activities, provided by participating brokerage firms and banks, before, during, and after they make transactions and flags suspicious behavior, such as a trader making a series of risky trades after frequently communicating with someone he or she does not usually talk to, or a trader celebrating to an unusual degree after making a successful transaction.
In the year since its launch, the BRCA Share database is now helping doctors give women insight into their risk for developing breast and ovarian cancer. Women who have particular variations of the BRCA1 or BRCA2 genes, which act as tumor suppressors, can have an elevated risk of developing these cancers, but it can be difficult to tell how high this risk is due to the large number of potential variants of these genes. The BRCA Share database provides doctors with access to data on 6,200 variants of the genes, allowing them to compare a patient’s particular genetic variation with ones known to increase the risk of developing cancer and plan treatment accordingly.
The Pennsylvania Department of Transportation (PennDOT) has partnered with Carnegie Mellon University to install short-range communication radios at intersections in Pittsburgh to share traffic signal information with connected and self-driving vehicles. The technology allows for two-way communication between vehicles and traffic signals, so that, for example, a traffic light could warn cars in an intersection that someone has run a red light or an ambulance could ensure its route has all green lights to reduce response times. PennDOT has installed the technology in 35 intersections thus far and will deploy it in 11 more this year.
Facebook has developed a new artificial intelligence system called DeepText that learns the meaning of text users post to provide contextually relevant information and services. For example, DeepText can identify language in chat messages that suggests a user needs a ride and offer to request an Uber, detect when users are trying to sell something on their Facebook page and automatically format the post to display the price and item description, as well as help quickly identify and remove abusive or harassing posts. DeepText is capable of analyzing thousands of posts per second in 20 different languages, with close to human levels of accuracy.
Researchers at North Carolina State University have developed a prototype system of sensors called the Health and Environmental Tracker (HET) to help asthma sufferers track their symptoms and predict the onset of asthma attacks, which could help them take preventative measures. HET relies on a wrist-worn device that tracks environmental factors and movement, an adhesive chest patch that monitors biometric data such as respiration and blood oxygen levels, and a spirometer, which measures lung function. By compiling data from all of these sources, corresponding software can identify environmental and physiological factors that trigger a patient’s symptoms and provide early warnings.
A team of geographers and artists at Carnegie Mellon University have developed Terrapattern, a visual search engine for satellite imagery that uses a type of machine learning called a convolutional neural network to allow users to rapidly search through images. Users can select a particular mapped image, such as an image of a football field or a bridge, and Terrapattern will provide a list of all similar locations based on their visual features. Terrapattern is still in early development but could have substantial humanitarian implications. For example, aid workers could use Terrapattern after an earthquake to quickly locate all of the damaged bridges or collapsed buildings in a particular area to guide disaster relief efforts.
DefinedCrowd, a company participating in the Microsoft Accelerator, a program to help later-stage tech startups scale their business, provided the first public demonstration of its ability to provide near-real-time processing of language data. Companies using machine learning techniques to interpret spoken and written language need good data to train their algorithms. DefinedCrowd partners with linguistics departments in universities around the world to hire students to analyze the data and provide additional information to help the machines better understand content. For example, workers might be asked to score whether a particular sentence is curious or sarcastic.
NASA has launched the Coral Reef Airborne Laboratory (CORAL), an initiative to use advanced imaging sensors to map four major coral reef areas—Hawaii, Palau, the Mariana Islands, and portions of the Great Barrier Reef—to provide baseline maps for future research. CORAL will fly over these areas using planes equipped with a special type of spectrometer, a device to measure particular bands of light, that can provide more detailed data than traditional satellite imagery, such as water depth and temperature. These maps will be valuable for conservation efforts by helping track the effects of climate change and other factors that kill off coral reefs, which can have a dramatic impact on ocean ecosystems.
The U.S. Department of Health and Human Services (HHS) has created a new website to publish detailed transparency information about health information technology (IT) products and services. HHS created the website to support disclosure rules ONC implemented in 2015 that require health IT developers to provide detailed, easy-to-understand information about the benefits and limitations of their products. In particular, HHS hopes this website will help discourage information blocking—the illegal but difficult to detect practice of restricting the interoperability of health data to discourage providers from switching vendors.
After Florida implemented its Prescription Drug Monitoring Program in 2011 to track individual drug prescriptions, the top prescribers of opioids in the state issued 6.2 percent fewer prescriptions and reduced the total volume of opioids prescribed by 13.5 percent after just one year. The statewide database allows doctors to review a patient’s past prescription history to see if he or she regularly seeks out opioid prescriptions from multiple doctors, which can be a sign of addiction or illegal activity.