Johns Hopkins Whiting School of Engineering 1996 Annual Report [text version]

Whiting School of Engineering 1996 Annual Report

Computer Science
To Lend a Helping Hand
Which Witch is Which?
New Center Makes a Point
Department Facts

To Lend a Helping ‘Hand’
The word “robot” has been part of our language and culture since 1921, when Czech playwright Karel Capek coined the term in his play R.U.R. (Rossum’s Universal Robots). Today, their potential to execute repetitive motions without tiring—and with a high degree of accuracy—have made them attractive to the manufacturing industry, and more recently, the medical field. Yet the computations required for robots to accomplish even the simplest task are mind-boggling, and computer scientists have become indispensable in their design and operation. Russell Taylor, a new professor of computer science in the Whiting School, is a leader in the field of computer-integrated surgery, which includes medical robotics.

After a distinguished career at the IBM T.J. Watson Research Laboratory, Taylor is continuing his work with several surgeons at the School of Medicine on the challenges they face in minimally invasive surgery (MIS) and aggressive cancer therapy. MIS procedures allow patients to suffer less pain and recover more quickly, but surgeons can lose some hand-eye coordination and dexterity due to the small incisions. Taylor and his colleagues have developed a telerobotic assistant that can actually perform simple tasks for the surgeon, allowing increased coordination and accuracy. In another project, the National Science Foundation has funded an effort to examine image guidance for localized liver cancer therapy. “The technical challenge here,” according to Taylor, “is placing a needle containing radioactive pellets very accurately on a soft tissue organ in the abdominal area—while the patient is breathing. Developing a system to do this requires solutions to a number of engineering research problems in robotics and image processing.” Taylor believes that a robotic assistant, guided by fluoroscopic X-ray images, can accurately and consistently place the pellets into a tumor in an optimized treatment pattern derived from CT (computed tomography) images. In fall 1995, he co-chaired the Second Annual International Symposium on Medical Robotics and Computer Assisted Surgery, held in Baltimore. Taylor, who received a bachelor of science degree from Hopkins in 1970, taught a new course sequence in Quantitative Medical Computing during the fall and spring semesters.

Which Witch is Which?
It took most of us quite a long time to learn English (or any other natural language, for that matter). From the time we were infants, we practiced until we knew intuitively the grammar rules that would allow us to communicate with others. Assistant Professors Eric Brill and David Yarowsky think that computers can also be taught to recognize a natural language, but it is something much easier said than done.

“The big problem we face,” says Brill, “is a knowledge acquisition bottleneck. Take the sentence: ‘we bought the car with a broken window.’ You and I have the cultural background to understand that the car has a broken window. The computer, being very naive, might reason that ‘a broken window’ is acceptable currency for car purchasing.” So it isn’t enough to give a computer words; it also needs linguistic and world knowledge.

One method tried by researchers was to ask, “What do I know about this language?” They fed their answers to that question, including grammar rules, into the computer. That approach has not worked well; language is much more complex than we realize, and the computer is truly a tabula rasa being.

Brill and Yarowsky use a new technique that employs corpus-based language processing. Using supervised learning, the method involves downloading a vast amount of text, in this case issues of the Wall Street Journal, and diagramming the text for the computer (remember high school English?). Brill’s work in this area addresses the parts of speech and phrase structure. Yarowsky’s research examines word sense disambiguation, in which two words sound the same but have different meanings, such as “rain” and “rein.”

The short-term applications of their work may result in improved spelling correction programs that could identify, for example, when “witch” was used for “which.” “In the long term, we may see the computer as a sophisticated language user, able to solve problems interactively with a human partner, summarize technical papers, and much more,” comments Brill.

New Center Makes a Point
In the highly-acclaimed animated film, The Point, protagonist Oblio lives in a world easily defined by points, lines, planes, and surfaces. The fact that Oblio’s head is curved rather than pointed is painfully obvious to his peers and causes great consternation. In the real world, the study of geometric objects of all types is fundamental to understanding the properties and boundaries of our physical surroundings. Computer scientists have taken the ancient theorems and proofs that define geometry and have applied them in modern ways. The relatively young field of geometric computing studies methods for computationally processing collections of geometric objects. Researchers use computer algorithms and software to study properties that include convexity, proximity, intersection, decomposition, and display. The area holds numerous application possibilities, from multimedia technologies and computer vision to robotics and astronomy. One example of a highly visible achievement in this field is Walt Disney’s Toy Story, the first feature-length film produced entirely through computer-generated rendering and animation of geometrically-defined mathematical models.

Recently, Professors Michael Goodrich and Rao Kosaraju won a grant from the Army Research Office (ARO) to establish a Center for Geometric Computing jointly with Duke University and Brown University. The Center’s focus is on applications to intelligent systems. As part of ARO’s multidisciplinary university research initiative, the grant gives computer science departments at the three institutions $1.5 million each over five years. The ARO made only 27 such awards in 14 focus areas. Those organizing the Center hope to “bring together computational researchers synergistically with scientists, engineers, and artists to drive the development of new geometric computing technologies....”

Of the new Center, department chair Gerry Masson says, “There will now be enormous opportunities for further growth and development in computer science in the educational and research aspects of geometric computing at Johns Hopkins and other institutions.”

Established 1986
The study of computer science was originally part of the Department of Electrical Engineering and Computer Science, which was established in 1981.

Phone 410-516-8577

Email comp_sci@jhu.edu

WWW http://www.cs.jhu.edu/

Students
1995-96 Academic Year
Graduate: 42
Undergraduate: 79

Faculty and Researchers
Gerald M. Masson, Chair
Yair Amir
Baruch Awerbuch
Eric Brill
Robert Cypher
Michael T. Goodrich
Smaragda Konstantinidou
S. Rao Kosaraju
Sudobh Kumar
Steven L. Salzberg
Scott Smith
Russell H. Taylor
Lawrence B. Wolff
David Yarowsky

Research Areas
Algorithm Design and Analysis
Artificial Intelligence
Computational Biology
Computer Graphics
Computer Vision
Distributed Computing
Fault Tolerant Systems
Geometric Computing
Medical Robotics
Networking
Parallel Computation
Programming Languages
Speech and Language Processing

Computer Science To Lend a Helping Hand Which Witch is Which? New Center Makes a Point Department Facts

Computer Science
To Lend a Helping Hand
Which Witch is Which?
New Center Makes a Point
Department Facts