Published on April 1st, 2016 | by Joshua New0
10 Bits: the Data News Hotlist
This week’s list of data news highlights covers March 26 – April 1, 2016 and includes articles about a machine learning system that can identify unlicensed money lenders and new technique that uses lasers to detect bacteria on food.
A partnership between the city of Chicago and insurance company Allstate has taken advantage of the company’s data analytics expertise to improve public services with insights from open data. Chicago has 15,000 food-serving establishments but just 42 health inspectors, so to make better use of the city’s limited resources, Chicago and Allstate analyzed 600 open data sets that could reveal known risk factors, such as time since the last inspection or unusual weather, to develop a model for more effectively prioritizing health inspections. Chicago and Allstate are taking a similar approach for elevator inspections to ensure inspectors are targeting elevators that pose the greatest public safety risks.
The Defense Advanced Research Projects Agency (DARPA) has launched the Spectrum Collaboration Challenge (SC2) to encourage the development of machine-learning systems that can allow wireless devices to collaboratively share radio frequency spectrum. By embedding these learning systems in wireless military and civilian devices, the number of which is rapidly increasing, device makers can ensure that the increased demand for spectrum does not lead to interference or reduced functionality. The challenge will run from 2017 to 2020, and the winning project will receive a $2 million prize.
Canadian company Maluuba has developed a deep-learning algorithm capable of reading and answering questions about text with a high level of accuracy. Maluuba trained its algorithm on hundreds of children’s stories and had it answer multiple-choice questions about the text, which it could answer correctly approximately 70 percent of the time. Unlike other text-comprehension algorithms, Maluuba’s algorithm still works when it only has a small amount of text to analyze. Maluuba hopes to use this approach to teach computers to comprehend mundane documents such as user manuals to distill important points so humans do not have to.
The Band of Russia is using a machine learning system developed by analytics company Yandex Data Factory to identify organizations illegally lending money, which frequently finance other illegal activity and scam consumers. The system analyzes thousands of websites related to loans and can correctly identify if an organization is licensed with 98 percent accuracy. The Bank of Russia has used this system to reveal 2,500 organizations suspected to be involved in unlicensed lending.
The World Bank and San Francisco-based data company Premise have been piloting a system to use crowdsourced smartphone photos to monitor price fluctuations of goods in areas where government statistical agencies lack the resources to collect sufficient economic data. People can take pictures of price tags as they shop and submit it to an app developed by Premise, which aggregates this data for a given area to better understand how prices are changing. With this data, governments that would not otherwise have the resources to do so can better understand the impact of economic policy decisions. The World Bank and Premise are conducting the pilot, which began a year ago, in 15 countries, including Cambodia, Ghana, and Brazil.
Microsoft has developed a pair of tools that use computer vision and natural language processing artificial intelligence software to help people with visual impairments better understand their surroundings. One of the tools, called Seeing AI, is an app that can be used with special glasses or a smartphone to provide users with a detailed audio description of their surroundings, such as the age, gender, and expressions of nearby people. The other tool, CaptionBot, is a prototype website that can provide detailed descriptions of user-uploaded images. The tools are part of Microsoft’s Cognitive Services project, which focuses on developing new apps and services powered by machine learning.
A joint initiative from Harvard University, the University of Oxford, and Dubai’s Museum of the Future called the Institute for Digital Archeology is using 3D-modeling and robotics to recreate a detailed replica of a Roman monumental arch from the second century destroyed by ISIS militants in Palmyra, Syria. The Institute for Digital Archeology compiled dozens of photographs of the arch to create a detailed 3D model, which robotic stone-carvers are using to create a 12-ton, 20-foot-tall scale replica in Carrara, Italy.
Researchers from Harvard University and the National Center for Atmospheric Research have developed a model for predicting heat waves in the eastern United States up to 50 days in advance by analyzing ocean temperature data. The researchers’ analysis revealed that extreme heat in the eastern U.S. have historically been preceded by patterns of anomalous surface temperature readings in the Pacific Ocean seven weeks in advance. By knowing when a heat wave is likely to occur, utility companies, farmers, and other industries and services impacted by extreme weather can better plan resiliency measures.
U.K.-based mobile navigation software company Citymapper has developed a new service called SuperRouter that can analyze costs, wait times, and routes for all methods of transportation, including public buses and trains, taxi services, walking, and cycling, to determine the fastest and most cost-effective routes possible. For example, SuperRouter can generate routes combining walking and a subway ride that will recommend substituting a particular portion of the subway ride with an Uber if a train is delayed.
Researchers at the Korea Advanced Institute of Science and Technology have developed a method to detect the presence of potentially dangerous bacteria on food using lasers and image analysis. The method involves shining a laser on a piece of food to analyze how the light scatters through it—when a piece of food has a large amount of bacteria on it, the microscopic movements of the bacteria will alter the results. By taking a rapid series of photographs and analyzing the subtle differences in the images, the researchers were able to identify the presence of bacteria. Though the technique cannot distinguish between different types of bacteria, it could serve as an effective precautionary measure to avoid foodborne diseases. Moreover, the equipment is cheap and simple enough to integrate into ordinary refrigerators.
Image: Jerzy Strzelecki.