Mining for answers

Dean's Award winner gets results connecting computer science and social work

Photo by Will Owens

Social workers know a lot about looking for a needle in a haystack. Accessing information can feel that way as they mine through endless case files to make connections and evaluate the quality and success of their programs.

In North Carolina, when the state’s 100 counties were charged with evaluating their Work First programs, each had to figure out how to share information more easily with state and federal officials, whose agencies fund the locally run programs.

This was just the opportunity Hye-Chung Kum was looking for.

A graduate student in computer science, Kum has a passion for social action. When she came to UNC-Chapel Hill from South Korea, she wanted to find a project that would put her computer science expertise to use in such an arena. In 1998, Kum became an integral part of a project in the Jordan Institute for Families that already has made life easier for numerous social workers across the state. In the process, she completed a Master of Social Work degree to complement her doctoral studies in computer science.

Working with Dean Duncan, research associate professor at the School of Social Work, and Database Manager Kimberly Flair, Kum developed a data mining system that facilitated the management and analysis of data from Work First programs. The project earned Kum a Dean’s Award during The Graduate School’s Centennial Celebration in March, 2003.

Her work, however, did not end there.

As a Paul Hardin Dissertation Fellow in the Royster Society of Fellows, she is continuing to bridge her passion for social action and her studies in computer science.

“With the Work First project I found that one of the questions that often comes up is ‘What are common patterns of services to clients?’” she said. “Clients are getting a group of services over time, so social workers want to know what the common patterns are, and that involves looking at things from an overview and bringing everything together.”

To answer this nagging question that many researchers have, Kum developed ApproxMAP, a new approach to study social information.

“Analyzing a sequence of sets is a difficult combinatorial problem that has gained much interest recently in data mining,” she said. “There was no satisfactory method available that could solve real problems, so I worked on designing a realistic method for sequential pattern mining. It builds on the work done in DNA sequencing, which analyzes a sequence of letters.”

Kum pointed out that, currently, many decisions about social work programs are based on ideas and impressions of the data rather than truly accurate information. Her new method organizes and summarizes a sequence of sets to discover the underlying trends or consensus patterns in the data.

For Kum, the link between computer science and social work is an obvious one, but bringing the fields together in her curriculum could have been a challenge. She credits the flexibility of The Graduate School and her doctoral program with allowing her to explore her unique interdisciplinary interests.

North Carolina’s front-line social workers are grateful, too.

- Alexandra Obregon

© 2004, The Graduate School, The University of North Carolina at Chapel Hill
All text and images are property of The Graduate School at the University of North Carolina-Chapel Hill. Contact Sandra Hoeflich at shoeflic@email.unc.edu to request permission for reproduction.