Association for Computing Machinery
Welcome to the May 13, 2016 edition of ACM TechNews, providing timely information for IT professionals three times a week.

Updated versions of the ACM TechNews mobile apps are available for Android phones and tablets (click here) and for iPhones (click here) and iPads (click here).


Google's Artificial-Intelligence Tool Is Offered for Free
The Wall Street Journal (05/12/16) Daniela Hernandez

Google is open-sourcing the code underlying a program that helps machines understand written English, as well as its accompanying software programming toolkit, for distribution or modification. The toolkit, SyntaxNet, helps machines deconstruct sentences and decode word-phrase relationships. Once sentences are parsed, the software can zero in on the most relevant words. Applications the software is used for include filling in Google's Knowledge Graph information repository, on which the Google search engine is founded. According to Google, the "Parsey McParseface" program can understand relationships between words in a sentence with more than 94-percent accuracy, which almost equals the performance of Google's human linguists. Among the system's limitations is a restriction to English and an inability to determine relationships between sentences, while longer sentences can muddle its accuracy. Experts also are uncertain how well it will work out-of-the-box in specialized settings such as healthcare and biotechnology, which require large volumes of data representing a field's particular vocabulary and its uses. Experts say natural language processing is essential to enhancing technologies such as virtual assistants, chatbots, automated language translation, decision-support systems, and relationship-revealing research tools.
View Full Article - May Require Paid Subscription | Return to Headlines | Share Facebook  LinkedIn  Twitter 

Altering a Robot's Gender and Social Roles May Be a Screen Change Away
Penn State News (05/11/16) Matt Swayne

Pennsylvania State University (PSU) researchers have found robots can keep their parts and still change their gender, as the arrival of robots with screens has made it easier to assign distinct personalities. They demonstrated that people found feminine cues on the robot's screen were enough to convince them a robot was female, according to PSU doctoral student Eun Hwa Jung. The findings could help robot developers customize robots for certain roles and to serve certain populations. "The screen, by itself, helped participants perceive whether the robot was male or female," Jung says. PSU professor S. Shyam Sundar says this shows it may not be necessary to alter the robot's shape or features to meet users' expectations and preferences. Participants in PSU's study assumed a robot without any gender cues was male, and also found male robots to be more human-like, more animated, and less anxious. "Gender is just one example that we tested here, but we see implications for other role definitions that we can potentially outfit a robot with by just manipulating a screen," Sundar says. The researchers presented their findings this week at the ACM CHI 2016 conference in San Jose, CA.

Intel Sees Mobile Future in Atomic Research
IDG News Service (05/12/16) Peter Sayer

Intel is clearing a path for future low-power mobile technologies with a new five-year collaboration with the French Alternative Energies and Atomic Energy Commission's (CEA) Laboratory for Electronics and Information Technology (LETI). The initiative will concentrate on basic development of faster wireless networks, secure low-power technologies for the Internet of Things, and three-dimensional displays. CEA managing director Daniel Verwaerde says although his organization will continue its mission to conduct nuclear weapons research with computer modeling, its areas of focus have broadened to include materials science, climate, health, renewable energy, security, and electronics. The last two areas will form the core of the new effort with Intel, as LETI researchers share information with Intel scientists. Intel's Rajeeb Hazra says his company is greatly reliant on overseas researchers. He notes Intel has more than 50 European labs, with four of them investigating exascale computing. Both CEA and Intel have not disclosed who will own the commercial rights to the collaboration's end products, but each emphasize their rights are protected. "It's a balanced agreement," says CEA Technology director Stephane Siebert.

Ingestible Origami Robot
MIT News (05/12/16) Larry Hardesty

Researchers at the Massachusetts Institute of Technology (MIT), the Tokyo Institute of Technology, and the University of Sheffield have demonstrated an origami robot that can unfold from a ingestible capsule and use external magnetic fields to navigate across the stomach wall to extract a swallowed button battery or repair a wound. "For applications inside the body, we need a small, controllable, untethered robot system," says MIT professor Daniela Rus. The robot propels itself via a "stick-slip" motion, in which its appendages cling to a surface through friction when it executes a move, but slip free again when its body flexes to change its weight distribution. To ensure this form of locomotion works, the robot uses a flexible biocompatible material instead of Mylar, as its predecessor did. The device also is composed of two layers of structural material sandwiching a material that contracts when heat is applied, while a pattern of slits in the outer layers determines how the robot will fold when the middle layer shrinks. The robot's design also incorporates a fin concept so it can propel water to move forward, while a permanent magnet responds to the external magnetic fields that guide it. The researchers tested the robot in a simulated stomach and esophagus.

Young Women in STEM Fields Earn Up to One-Third Less Than Men
Ohio State University (05/10/16) Jeff Grabmeier

Women with Ph.D.s in science and engineering fields earn 31 percent less than men one year after they graduate, according to a study led by Ohio State University professor Bruce Weinberg. The pay gap was 11 percent when researchers took into account that women tended to graduate with degrees in fields that generally pay less than fields in which men got their degrees. Men were more than twice as likely to complete dissertations in lucrative fields such as engineering and were 1.5 times more likely to study computer science, math, or physics. In addition, although the private sector tends to pay the largest salaries, women were more likely than men to work in government and academia. The rest of the pay gap disappeared when the researchers controlled for whether women were married and had children. The importance of helpful family policies is supported by the fact that single and childless women tended to have less of a pay gap than those who were married and those who had children. The researchers had data, not previously available to scientists, on 1,237 students who received Ph.D.s from four U.S. universities from 2007 to 2010 and were supported on research projects while in school.

Paper Gets 'Smart' With Drawn-on, Stenciled Sensor Tags
UW News (05/11/16) Michelle Ma

Researchers from the University of Washington (UW), Disney Research, and Carnegie Mellon University have developed a method in which small, off-the-shelf radio-frequency (RFID) tags are attached, printed, or drawn onto paper to create interactive interfaces. The interfaces can enable the paper to respond to gesture commands and link to the digital world. The PaperID technology employs RFID tags with individualized IDs that can be detected by readers in the same room, and whose antennas can be stenciled on paper with conductive ink. In response to the disturbance of the signal path between the reader and tag caused by a person's gestures, algorithms can identify the movements and classify the interruption as a specific instruction. "These little tags, by applying our signal-processing and machine-learning algorithms, can be turned into a multi-gesture sensor," says UW doctoral student Hanchuan Li. The researchers found specific RFID tag applications are more conducive to certain desired interactions, such as tags stuck to paper for on/off button commands. "If RFID tags can make interfaces as simple, flexible, and cheap as paper, it makes good sense to deploy those tags anywhere," Li says. The project was presented this week at the ACM CHI 2016 conference in San Jose, CA.

Big Thinking in Small Pieces: Computer Guides Humans in Crowdsourced Research
CMU News (05/09/16) Byron Spice

Carnegie Mellon University (CMU) researchers have demonstrated that computer-guided non-experts can collectively research and craft coherent reports, using a system called the Knowledge Accelerator. The system employs a machine-learning program to sort and structure information exposed by individuals focused on just a small segment of a larger project. It uses those findings to make new assignments and assembles an architecture for the final report based on its emerging comprehension of the topic. CMU professor Niki Kittur says the Knowledge Accelerator keeps track of the big picture so the various contributors to the project can crowdsource the research. He and fellow researchers at CMU's Human-Computer Interaction Institute developed Alloy, a system that uses machine learning to inject structure and coherence into human-collected information, and which automatically identifies patterns and categorizes the information. The system was asked to research and write reports based on 11 questions, and assessors determined the Knowledge Accelerator output to be superior to comparable Web pages. The Bosch Research and Technology Center's Ji Eun Kim says the center is extending the tool for use in diagnostic and repair services on the strength of its knowledge synthesis capability. The researchers presented the Knowledge Accelerator this week at the ACM CHI 2016 conference in San Jose, CA.

Algorithm Clones Van Gogh's Artistic Style and Pastes It Onto Other Images, Movies
Technology Review (05/10/16)

University of Freiburg researcher Manuel Ruder takes famous works of art and transfers their style to a range of video sequences taken from movies and TV programs, resulting in a new rendering of the videos and the possibility of doing them in almost any style. The technique relies on deep neural networks, which consist of many layers that each extract information from an image and then pass on the leftover data to the next layer. Last year, University of Tubingen researchers began studying artistic style in this way, and found it is possible to capture artistic style by examining the correlations between layers, instead of the information in each layer. Ruder notes the key discovery from the Tubingen research was the content of an image can be separated from the artistic style. The Tubingen researchers also found they could copy and paste the artistic style onto the content of any other image. Applying this technique to video led to the discovery that small differences between successive frames can lead to big differences in the way the new artistic style is used, making the video appear jumpy and visually incoherent. The Freiburg researchers addressed this problem with an algorithm that analyzes the deviations between successive processed frames.

Smartphones Uncover How the World Sleeps
University of Michigan News Service (05/06/16) Nicole Casal Moore

A new study led by University of Michigan researchers helps quantify the role society plays in influencing sleeping patterns. The team used math modeling, mobile applications, and big data to produce the research. Several years ago, the team released an app called Entrain, which helps travelers adjust to new time zones by recommending custom schedules of light and darkness. The quality of the app's recommendations depended on the accuracy of the travelers' information, and the researchers say this motivated users to be particularly careful in reporting their lighting history and sleep habits. The team analyzed sleep data from thousands of people in 100 nations for patterns in what amounted to a circadian rhythm simulator. A mathematical model, the simulator is based on the field's deep knowledge of how light affects the brain's suprachiasmatic nucleus. The researchers could dial the sun up and down in the model to see if the correlations still held in extreme conditions. They say the work also demonstrates mobile technology can be a reliable way to gather massive datasets at very low cost.

The Internet of Drones Is Coming
Motherboard (05/08/16) Michael Byrne

Drones are advanced enough to conceivably be used for delivering goods, but a significant barrier remains in the form of drone traffic density, and Robert J. Hall at AT&T Labs believes the answer lies in an "Internet of drones." He says drones can best avoid each other if they know about each other, which is the essence of his Geocast Air Operations Framework (GAOF) prototype. "The goal of the work is to demonstrate a path toward an improved system for the operation of drones, with the necessary secure command and control among all legitimate stakeholders, including drone operator, [U.S. Federal Aviation Administration], law enforcement, private property owners, and citizens," Hall says. GAOF works by automatically flipping between cellular and wireless ad-hoc network tiers depending on availability. The system also makes use of geographic addressing (GA), which is similar to how subnets work on the Internet. In GA, circles centered around different latitudes and longitudes are assigned their own addresses, which are shared among all drones within that circle. Every device that wishes to monitor an area comes with up a query message, which is then transmitted to a specific geographic address. The drones within that address region send their replies back to the geographic address of the querying drone.

Do You Have a Secret? The Way You Write Emails May Give It Away
New Scientist (05/05/16) Aviva Rutkin

A study of 61 people with highly sensitive secrets participated in an experiment at the University of Maryland (UMD) in College Park to determine how well their emails could conceal that knowledge. UMD's Yla Tausczik and colleagues were given access to the participants' email accounts, the date when their secrets began, and the people they were hiding the secret from or sharing it with. Once identifying information was removed, the researchers sifted through the language used in more than 59,000 messages, and text-analysis software showed secret-keepers were "hypervigilant" in terms of social interaction, according to Tausczik. They especially were careful to maintain relationships with those they wanted to keep secrets from, sending them more emails each month after they started keeping a secret than they had before. The researchers also studied participants' relationships with those who knew their secret, as secret-keepers tended to emulate the language of their confidants much more, and use more negative emotional words and terms related to insight or causation. Stanford University's David Markowitz says the research demonstrates how secret-keepers attempt to hide their secrets by attempting to act normally with those they are keeping in the dark. University of California, Santa Barbara's Norah Dunbar says a long-term goal of such work could be the refinement of automatic and predictive deception-detecting systems.
View Full Article - May Require Free Registration | Return to Headlines | Share Facebook  LinkedIn  Twitter 

Tech Helps Teens Battle Asthma
University of Rochester NewsCenter (05/06/16) Bob Marcotte

Hyekyun Rhee, chair of nursing science at the University of Rochester's (UR) School of Nursing, has developed ways to help teens with asthma better manage their condition. She enlisted the help of UR professor James Allen to devise the Mobile Phone-Based Self-Management Aid for Adolescents, which is designed to understand texting terminology, as well as words that would indicate symptoms, activities, or medications. In a pilot study, teens and parents who used it in a two-week trial reported better asthma self-management by teens and a better teen-parent partnership in managing the condition. Rhee also worked with UR professor Mark Bocko on the Automated Device for Asthma Monitoring, which continuously monitors coughs 24 hours a day in conjunction with activity levels using a built-in accelerometer. The data can be downloaded to mobile devices or a computer to see daily, weekly, or monthly patterns of symptoms and adjust medication or behavior. Rhee also has received a $3-million grant from the U.S. National Institutes of Health to lead a randomized, controlled study to test the Peer-led Asthma Self-Management for Adolescents Program in Buffalo, Baltimore, and Memphis.

USC/JPL Scientist Made Big Data Stories Possible on the Panama Papers
USC News (05/09/16) Tarika Lall

University of Southern California (USC) professor and Jet Propulsion Laboratory researcher Chris Mattmann helped create the open source Apache Tika software used to sift data from the Panama Papers. In an interview, Mattmann says Tika was a USC postdoctoral project he developed with Jerome Charron based on the understanding that many different programs are required to identify file types and extract data. "Tika is a 'digital Babel fish:' you can throw any file at it, and Tika will give you an understanding of the content inside," Mattmann notes. Organizations such as the U.S. Defense Advanced Research Projects Agency and the U.S. National Aeronautics and Space Administration have invested in Tika's evolution "into a software that can extract more complex data about people, places, dates, and times from any file, including images and videos," he says. Tika can identify people and their whereabouts from a Word file, and machine translation also is being incorporated to enable not only language identification, but automatic translation into desired languages as well. Mattmann says Tika does not breach privacy because it does not collect people's data by itself. However, he acknowledges "it can be used for questionable purposes...but we can't control that." Mattmann imagines Tika could be used for metadata analysis, as well as boosting the performance and efficiency of search engines and content management systems.

Abstract News © Copyright 2016 INFORMATION, INC.
Powered by Information, Inc.

To submit feedback about ACM TechNews, contact: [email protected]
Current ACM Members: Unsubscribe/Change your email subscription by logging in at myACM.
Non-Members: Unsubscribe