Frederick Jelinek
Frederick Jelinek (1932-2010) was a notable figure in the field of information theory, recognized for his pivotal contributions to modern computer language translation and speech recognition technologies. Born Bedřich Jelínek in Kladno, Czechoslovakia, Jelinek faced significant challenges during his early life, including the impact of the Holocaust on his family. After immigrating to New York City in 1949, he pursued engineering studies at City College and later earned a doctorate from the Massachusetts Institute of Technology (MIT).
Jelinek joined IBM in 1972, where he led the Continuous Speech Recognition Group for over two decades. His innovative approach to speech recognition relied on statistical methods, allowing computers to process natural language more effectively than previous models that attempted to mimic human cognitive processes. After retiring from IBM, he became the director of the Center for Language and Speech Processing at Johns Hopkins University, continuing his research and mentoring students.
Throughout his career, Jelinek received numerous honors, including election to the National Academy of Engineering. He remained connected to his roots in Czechoslovakia, contributing to the development of speech science there after the fall of communism. Jelinek's legacy is marked by his unique blend of engineering and linguistics, shaping the future of human-computer interaction.
Subject Terms
Frederick Jelinek
Developer of speech recognition technology
- Born: November 18, 1932
- Place of Birth: Kladno, Czechoslovakia (now Czech Republic)
- Died: September 14, 2010
- Place of Death: Baltimore, Maryland
- Primary Company/Organization: IBM
Introduction
Frederick Jelinek was a pioneer of information theory whose work created the foundation of modern computer language translation and speech recognition technology. Jelinek and colleagues at IBM took a novel approach to computer language processing, based on the specific strengths of the computer rather than trying to imitate human thought; the statistical methods they developed proved crucial to enabling computers to transcribe, understand, and translate natural human speech.

Early Life
Frederick Jelinek was born Bedřich Jelínek in 1932 to a Jewish family in Kladno, Czechoslovakia, a city near Prague; his father was a dentist. Jelinek's early education was sporadic and unconventional: Because of his Jewish heritage, he was barred from attending local public schools and instead attended makeshift classes organized by Jewish members of the community. Jelinek recalls the classes having an ever-changing roster of teachers, as members of the Jewish community were sent off to concentration camps. Even that formal education ceased in 1942, when the Nazis forbade any instruction for Jews. Jelinek's father, who was Jewish, died in the Terezin concentration camp, but his mother, a convert to Judaism, was allowed to remain free, as were her children, who were considered to be half Jewish. This allowed them to survive the Holocaust, and Jelinek resumed his education in Czechoslovakia. However, because of his lack of formal education, he did poorly in the gymnasium (the equivalent of a college preparatory high school).
In 1949, Jelinek immigrated with his mother and sister to New York City, where he found the opportunity to demonstrate his academic abilities. Although his original goal was to be a lawyer, after graduating from high school he took courses in electrical engineering at City College, then attended the Massachusetts Institute of Technology (MIT) with financial assistance from the National Committee for a Free Europe. He received his doctorate in engineering from MIT in 1962 and joined the faculty of Cornell University.
Life's Work
Although Jelinek has said that he entered the field of engineering out of expediency rather than aptitude or personal interest, his studies in that field allowed him to begin his work in information theory and gave him the mathematical background for his pioneering work in computer translation and speech recognition. He studied at MIT in the years 1954–62, a period in which MIT was also a center of research into information theory. Among the faculty members at MIT during this time were Claude Shannon, who essentially created the field of information theory; Robert Fano, who developed the Shannon-Fano coding system in collaboration with Shannon; and Peter Elias, who developed convolutional codes, a type of code used in data transfer. Jelinek was also exposed to the ideas of some of the most important linguists of the day while at MIT, including the Russian linguist and literary theorist Roman Jakobson, who pioneered the structural access of language, and Noam Chomsky, who developed the theory of transformational grammar. Jelinek became so interested in linguistics and information theory that he considered leaving the field of engineering, but he was convinced by his thesis adviser, Fano, that he should finish his Ph.D. studies in electrical engineering and could shift his focus afterward.
Jelinek's first faculty position, which he accepted in 1962, was at Cornell University; he was attracted to the university in part by the opportunity to work with the linguist Charles F. Hockett in applying information theory methods to linguistics. Although that collaboration did not work out, Jelinek remained at Cornell for ten years, working in the field of information theory. In 1972, he applied for a summer position at the Thomas J. Watson Research Center at IBM, where research on speech recognition was just beginning.
With this position, Jelinek began his first work in computer speech recognition. When the group's leader, Joe Raviv, accepted a job in Israel, Jelinek was promoted to the head of the Continuous Speech Recognition Group at IBM; he remained with the company for twenty-one years, until his retirement in 1993. Jelinek's group at IBM consisted primarily of Ph.D.'s in fields such as physics and information theory rather than linguistics or another speech-related subject; this allowed them to take a completely novel approach to speech recognition. The current approach was based on teaching the computer to process speech in a manner similar to that used by human beings. To allow a computer to follow this approach, it was necessary to break speech down into its smallest identifiable units, called phones, compare the sequence of phones with known word pronunciations, and accept as correct a transcription of the speech that best matched recognized sequences of phones. Jelinek's group took a different approach, one that capitalized on the specific capabilities of computers, that is, the ability to store large amounts of data and perform rapid computations. They decided that it would be impossible to specify an adequate series of rules for this process to succeed and instead applied mathematical and statistical methods that form the basis of modern speech recognition and transcription programs.
After retirement, Jelinek became the director of the Center for Language and Speech Processing at Johns Hopkins University, where he also held the position of Julian S. Smith Professor of Electrical Engineering; he worked at Johns Hopkins for nearly two decades, until his death in 2010. At Johns Hopkins, Jelinek continued to work closely with students and faculty, leading a series of summer workshops that brought together students and professionals from a variety of fields and professions to work on speech and language-processing problems.
Jelinek continued to visit Czechoslovakia as an adult, and after the fall of communism in that country in 1989, he initiated relationships with Czech researchers working in speech science and linguistics. He was instrumental in persuading IBM to set up a computer center at Charles University in Prague, taught in Czechoslovakia, and brought Czech scientists to work with his group at IBM. In this action he was fulfilling a promise made when he accepted funding from the Committee for a Free Europe, that he would help rebuild Czechoslovakia once it was no longer a Communist country.
Jelinek received an honorary doctorate from Charles University in 2001. He also received many other honors, including election to the National Academy of Engineering, being named one of the first twelve Fellows of the International Speech Communication Association, being named a Fellow of the Institute of Electrical and Electronics Engineers, and induction into the National Academy of Engineering in 2006.
Personal Life
Jelinek died of a heart attack while conducting research on the Johns Hopkins campus. He had been married since 1961 to Milena Jelinek, a Czech filmmaker, screenwriter, and professor of film studies whom he had met in 1957 on a visit to Czechoslovakia; they applied for a marriage license several times but were denied by the Czech government, and they could not be married until Milena was granted permission to visit the United States. The marriage produced a daughter, Hannah Sarbin, and a son, William. Jelinek was also survived by his sister, Susan Abramowitz; his half sister, Hirina Hlavac; and his grandchildren, Alex Sarbin, Sophie Jelinek, and Benjamin Jelinek.
Jelinek once commented that his career was largely a result of luck and circumstances and that he originally intended to be a physician, then a lawyer, rather than an engineer. He chose engineering in part because the course of study was only four years and in part because he felt his accent would be less of a hindrance in engineering than in law.
Bibliography
“Frederick Jelinek 1932–2010." National Academy of Engineering, 2024, www.nae.edu/189634/FREDERICK-JELINEK-19322010. Accessed 14 Oct. 2024.
Geller, Tom. “Talking to Machines.” Communications of the ACM 55.4 (2012): 14–16. Print.
Jelinek, Frederick. “The Dawn of Statistical ASR and MT.” Computational Linguistics 35.4 (2009): 483–94. Print.
Jelinek, Frederick. “Some of My Best Friends Are Linguists.” Language Resources and Evaluation 39.1 (2005): 25–34. Print.
Jelinek, Frederick. Statistical Methods for Speech Recognition. Cambridge: Massachusetts Institute of Technology, 1997. Print.
Lohr, Steve. “Frederick Jelinek, Who Gave Machines the Key to Human Speech, Dies at 77.” New York Times, 24 Sept. 2010, www.nytimes.com/2010/09/24/business/24jelinek.html. Accessed 14 Oct. 2024.