By Martin Weil
Monday, September 27, 2010; 6:17 PM
Frederick Jelinek, 77, a brilliant and determined engineer who overcame the challenges of a boyhood in Nazi-occupied Europe to come to America and teach computers how to recognize the sounds and sense of spoken English, died Sept. 14 in Baltimore.
Dr. Jelinek was at work at Johns Hopkins University when he died of what appeared to be a heart attack, said his son William. He spent much of his career at IBM's Watson Research Center in Yorktown Heights, N.Y., and in 1993 joined Hopkins, where he became a professor of electrical engineering and director of the Center for Language and Speech Processing.
A native of what is now the Czech Republic, Dr. Jelinek was brought to this country by his mother after his father died of disease in a Nazi concentration camp.
He obtained a doctorate from MIT in electrical engineering, made a specialty of information theory, and was credited as a pioneer in enabling computers to understand English.
"Fred's methods are at the heart of all systems today," one of his colleagues, Jason Eisner, told a Johns Hopkins publication about Dr. Jelinek's speech-recognition work.
In an interview, Eisner, an associate professor of computer science at Hopkins, said the recognition systems involve ordinary computers. Dr. Jelinek's "fundamental contributions were at the level of the mathematics and the algorithms" that made it possible for computers to determine what is being said to them and to respond.
Early efforts to do this centered on stocking computers with all the rules of linguistics and inserting information into their memories about all the ways a meaningful sentence could be assembled, said another colleague, Sanjeev Khudanpur, a Hopkins associate professor of electrical and computer engineering.
He said Dr. Jelinek chose another line of attack, proposing, in effect, to provide the computers with "lots and lots of text" - so much that they could infer where sentences probably were headed through speedy analysis of word context and phrase frequency.
Through "statistical magic," said Eisner, computers could be made to distinguish between the almost identical sounds of "wreck a nice beach" and those of "recognize speech."
After running through enough text, they could decide on the likely fourth word in a series after hearing the first three.
The field of artificial intelligence, in which Dr. Jelinek made major contributions, often seems to skirt the boundaries of science fiction and fantasy. Aspects of his own life own seemed suited to fiction. At the least, it was as he once said, "full of detours and compromises" as well as "lucky accidents."
He was born in Kladno, a town near Prague, on Nov. 18, 1932. His father, a dentist and physician, was Jewish and prepared to take the family to England to escape the growing Nazi menace. But he chose his patients' welfare at the last minute and remained behind, Dr. Jelinek's son said.
Frederick Jelinek's formal schooling was halted after the second grade. After the war, he found himself three grades behind others in his age group, and said in a speech that "during my entire studies I never received a single A in any subject."
In the late 1940s, fleeing from growing communist power, his mother, "by enormous effort," got the remnants of the family to America.
In addition to his son, of Briarcliff Manor, N.Y., Dr. Jelinek is survived by his wife, Milena Jelinek, of New York; his daughter Hannah Sarbin, of Larchmont, N.Y.; a sister; a half-sister; and three grandchildren.
Dr. Jelinek finished high school and then worked in a factory in New Jersey to help his struggling family.
He had wanted, he said, to "defend the unjustly accused" as a lawyer. But deciding that it would take too long to get a law degree, he enrolled in night classes in engineering at the City College of New York, despite "no previous inclination" toward the subject. In time, he got a doctorate from MIT, worked in information theory at Cornell University and began a research career at IBM.
At IBM, visitors included Nobel laureates to whom his speech-recognition group was shown as an "avant-garde" curiosity.
Occasionally, those visits were useful. One Nobelist, he said, even offered instruction on tying shoelaces.
After learning the new shoe-fastening technique, he said, he had "to admit that they seldom" came loose again.