IBM researcher Solomon Assefa likes to operate at the intersections of scientific domains and social institutions. It allows him to envision ways of using cutting-edge technologies to tackle new challenges. In a recent example of how his approach can pay off, he made the connections that resulted in a small team of IBM scientists helping boost a United Nations Children’s Fund social networking project that could improve the lives of millions throughout the continent of Africa.
The UNICEF project, called U-report, got its start as a text-messaging system for gathering information from young people in Uganda. Initially, UNICEF surveyed people via their mobile phones to get information about what was going on in their lives. The flow of messages turned into a torrent when youngsters, unsolicited, started reporting all sorts of problems in their communities–everything from dry wells to child abuse. More than 240,000 people signed up for the program. UNICEF’s staff couldn’t handle the deluge of messages.
This is where IBM’s technology came in. A team at the lab in Yorktown Heights, N.Y., took sophisticated machine learning and text analytics technology that had been developed for a different purpose and adapted it for U-report. Installed in Uganda in February, the technology is helping UNICEF size up problems so it can respond appropriately. Members of the Ugandan parliament get frequent U-report updates from UNICEF, called National Pulse. And, now, the U-report system has been expanded to several other African and Middle Eastern countries, including Zambia, Burundi and South Sudan, and, soon, the Democratic Republic of Congo.
Solomon Assefa is credited by his IBM Research colleagues with planting the seeds for the IBM/UNICEF partnership. Solomon, who joined IBM Research in 2004 after receiving his PhD. in physics and computer science from MIT, focuses on nanophotonics–the science of using fiber optics to transmit data at high rates of speed within computing systems. But he has broad interests. A native of Ethiopia who came to the United States for university studies, he wants to contribute to improving the lives of the world’s less fortunate people. That avocation landed him at a UNICEF conference on infant mortality in New York City in 2012, where he sat near one of the UNICEF staffers who was managing the U-report program and learned about how the organization was struggling to manage the large number of incoming text messages.
It took a while for Solomon to track down the right people within IBM who could help solve the problem, but, eventually, his search led him to Rick Lawrence, manager of the machine learning group. Two members of Rick’s team, Prem Melville and Vijil Chenthamarakshan, went to work on the project and developed a system combining text analytics and machine learning.
Here’s how their technology works: They built a text classification system that sizes up SMS messages as they’re gathered and stored in the U-report database. They based the classifications on a structure created by volunteer message reviewers–12 topics ranging from water, health and nutrition and violence against children to education, employment and emergencies. An important element in the technology is their “dual supervision” training technique–using both the historical human classification data and a machine classifier based on analysis of key words in the texts.
U-report has been quite successful. The system alerted public health officials to an ebola outbreak. Young people with concerns about HIV/AIDS are directed to nearby clinics for confidential testing. Messages about flooding in one of Uganda’s districts allowed UNICEF to respond quickly with assistance.
The system also provides feedback to government officials about the effectiveness of some of their social programs. In one case, after participants criticized the government’s Youth Fund, a program designed to help disadvantaged young people get funding for micro-enterprises, an investigation revealed that a design flaw in the program excluded many of the intended beneficiaries. That flaw was quickly remedied.
The technology work has been recognized in the external data mining community. A joint IBM UNICEF paper, “Amplifying the Voice of Youth in Africa via Text Analytics,” received the Best Paper Award in the Industry/Government Track at the upcoming ACM SIGKDD Conference on Knowledge Discovery and Data Mining.
The IBM Research team is now working on additional features that will make it possible for the system to assess the seriousness and urgency of text messages, so UNICEF and governments of African countries can prioritize their responses.
For Solomon, U-report is a first step back towards Africa–the land of his youth. He was also a key participant in the first strategy for launching the IBM Research Lab in Kenya, IBM’s first on the African continent. Solomon believes that multidisciplinary and multistakeholder collaborative initiatives will be crucial to Africa’s future. “A lot of innovation happens at the intersections of fields. We have to broaden our scope and get out of our corporate zones,” he says. “There’s a tremendous potential on the African continent due to economic growth and demographic shifts. We have to be partners and enablers in this historic transformation.”
This post originally published on asmarterplanet.com
If you want to learn more about the era of cognitive computing, download a free chapter of Smart Machines, a book by IBM Research Director John E. Kelly III, at the Web site of Columbia University Press, http://cup.columbia.edu/static/cognitive.