10 Bits: the Data News Hotlist
This week’s list of data news highlights covers June 25 – July 1, 2016 and includes articles about new projects supporting Vice President Biden’s Cancer Moonshot, as well as a new data mining technique that revealed why people make mistakes.
The White House has launched the Data-Driven Justice Initiative (DDJ), a coalition of 67 city, county, and state governments committed to using data to reduce unnecessary incarceration. The DDJ will use data-driven strategies to proactively provide health-care and social services to individuals at risk of recidivism, train and equip first responders to better respond to people experiencing mental health crises, and develop and deploy risk-assessment tools to better inform pretrial release decisions. The White House will provide DDJ participants with a variety of resources to support implementation of these strategies, including best practices for handling health and criminal justice data and funding evidence-based intervention tools.
GeoPlatform.gov, the U.S. federal government’s open data portal for geospatial data, is implementing a series of improvements to make geospatial data more accessible and usable. GeoPlatform.gov pulls geospatial datasets from Data.gov and provides users tools to analyze and build maps with this data. The site will soon provide a performance dashboard that monitors the quality and availability of the site’s data. Additionally, the GeoPlatform.gov team is developing tools to allow the site to suggest revelvant data layers or similar maps to users, as well as improve GeoPlatform Marketplace, which allows users to discuss their projects and data, share data, and create partnerships.
Toronto startup Whirlscape has developed Dango, a smartphone app that uses machine learning to analyze users’ text messages as they write them to automatically suggest relevant emojis. Whirlscape trained Dango’s algorithms on posts on Instagram, Reddit, and Twitter to learn which emojis correspond with different phrases. For example, typing “she said yes!” will prompt Dango to suggest the wedding ring and bride emojis.
Google has updated its Earth and Maps tools with new data from the Landsat 8 satellite, run by NASA and the U.S. Geological Survey, to create higher resolution and more accurate satellite maps. Previously, Google used data from the older Landsat 7 satellite and had to use software to fill in small gaps in this imagery, resulting in lower-resolution or duller maps. Landsat 8, which launched in 2013, provides more complete, detailed images with more accurate colors and captures these images twice as frequently as Landsat 7.
Vice President Joe Biden has announced new initiatives to support the Cancer Moonshot, an initiative to advance treatment and understanding of cancer with precision-medicine techniques and data sharing. To support the Cancer Moonshot, a host of federal agencies including the National Cancer Institute, Department of Energy, Department of Veterans Affairs, and the Food and Drug Administration will implement a variety of programs to accelerate progress, including: expanding researchers’ access to cancer data, making clinical research trials more accessible to cancer patients, using supercomputing to analyze cancer and pharmaceutical data, expediting the development and regulatory review process for novel cancer treatments, and fast-tracking the review process for cancer treatment-related patents.
Image-focused social media site Pinterest has developed a visual search tool powered by machine learning that allows users to easily identify objects they like in pictures and find it for sale online. With the tool, users can take a picture of an object in the Pinterest app, such as a watch, and it will search other Pinterest posts for pictures of that watch and notify users if those items are available for purchase. Pinterest will make the tool available to users within the next few months.
The U.S. Defense Advanced Research Projects Agency (DARPA) has passed an important milestone for its OrbitOutlook program, which tracks the hundreds of thousands of pieces of debris, old equipment, and other waste orbiting around the Earth. OrbitOutlook now incorporates live data streams from seven “space situational awareness” providers, which track debris orbiting the earth with sensor networks. The debris ranges from small paint chips to discarded rocket parts and can be incredibly dangerous to satellites given the high speed at which they orbit the Earth. DARPA is also developing algorithms that can automatically analyze this live data as well as validate data from non-certified sources to increase the amount of insight OrbitOutlook can provide military and commercial satellite operators.
Researchers from Microsoft, Cornell University, and Harvard University have developed a data mining technique that provides insight into the factors that contribute to poor decision-making in humans. The researchers analyzed a database of 200 million recorded online chess games and measured the impact of factors such as time pressure, skill level, and difficulty of decision and whether or not a player performed a good or bad move in the game, revealing their interesting relationships with decision-making. For example, decisions made in under approximately 10 seconds increases the likelihood of a bad move, and counterintuitively, a high skilled player can actually be more likely to make poor choices than a less-skilled player.
Dutch telecommunications company KPN has announced that it has finished deploying its long range (LoRa), low data rate network covering all of the Netherlands to support the Internet of Things. The network already provides coverage for 1.5 million connected devices, which do not need to transmit large amounts of data, but need to transmit and receive information in remote areas. KPN is also supporting three pilot projects with the LoRa network, including networked depth sensors in the port of Rotterdam, connected railway sensors in Utrecht, and a connected baggage handling system at Amsterdam’s Schiphol airport.
Researchers at the University of Georgia, the Massachusetts Institute of Technology, and the Max Planck Institute for Informatics have developed a smartphone app called GazeCapture to crowdsource data about how users look at their phones to help develop eye-tracking tools for mobile devices. GazeCapture uses smartphones’ front-facing capture to record where a user looks in a test that displays dots at different points on the screen, which provides researchers with useful data about how people interact with their phones outside a laboratory setting. GazeCapture has 1,500 users so far, and by collecting more data, researchers expect to be able to improve the accuracy of eye-tracking algorithms to the point where eye movements could be used to interact with smartphones as well as support medical diagnostic applications.