Improving the accuracy of Internet searches for images led UT Dallas doctoral student Yohan Jin to collaborate with a near-and-dear, if unexpected expert — his father.

Jin, a computer science student, studied ways to rule out “noisy” (or unrelated) keywords that are often next to images on the Internet. Refining words searches by mathematically slicing out unrelated text—such as words in captions—helps hone searches and improve their accuracy.

“It’s an important research problem because the text surrounding Web images is a good resource for identifying the images, but there are many noisy keywords as well,” Jin said. “This paper is the first to apply a graph approximation algorithm framework to the multimedia retrieval problem.”

A skilled computer scientist, Jin realized in the course of his research that his father could contribute a wealth of math expertise to the award-winning paper, along with the study’s other two authors from UT Dallas. The team members’ work netted them the “Best Paper Award” at the workshop on Semantic Learning Applications in Multimedia, held at the 2008 IEEE Conference on Computer Vision and Pattern Recognition.

“Actually, I didn’t know details about my son’s research until last winter,” said Kibum Jin, a professor at Soongsil University Computer Institute in Seoul, South Korea. “However, when my son talked to me in person about his idea for this paper, I could see a few places where I could make a contribution. It was a great experience to produce a research paper together.”

Jin’s graduate adviser and study co-author, Dr. Balakrishnan Prabhakaran, associate professor of computer science, said working with the father/son duo was interesting and helped strengthen the research.

“Yohan has been a great asset for my lab, and I am proud of his achievement,” Prabhakaran said. “When Yohan told me his father gave him ideas for this paper and asked if he could have his father as a co-author, I told him ‘Of course, by all means.’ I hope I can meet his father one of these days and tell him how proud we are of Yohan.”

Yohan Jin is positive and forward-looking about working closely with his father.

“I was excited to discover that there were some mathematic areas where I could use my father’s expertise,” Jin said.

“Having his name on my Ph.D. research paper is important and memorable for me. After this, I hope we can do even more work together.”

Jin plans to defend his dissertation in August and then join a data-mining team at MySpace.com. His co-author father couldn’t be more pleased.


Media Contacts: Brandon V. Webb, UT Dallas, (972) 883-2155, brandon.webb@utdallas.edu
or the Office of Media Relations, UT Dallas, (972) 883-2155, newscenter@utdallas.edu


Father and son: Kibum Jin (left) and Yohan Jin

 

Best Paper Award

Workshop on Semantic Learning
Applications in Multimedia
2008 IEEE Conference on Computer Vision and Pattern Recognition

“The Randomized Approximating Graph Algorithm for Image Annotation Refinement Problem”

Authors:
Yohan Jin, UT Dallas
Kibum Jin, Soongsil University Computer Institute
Latifur Khan, UT Dallas
B. Prabhakaran, UT Dallas