Fred Jelinek

Frederick Jelinek died, peacefully and unexpectedly, on 14 September 2010. Over a distinguished career of nearly fifty years, Fred made important contributions in areas ranging from coding theory and speech recognition to parsing and machine translation. But more important than these specific contributions was his role in persuading the fields of speech and language engineering to adopt statistical methods and the “noisy channel model,” returning to the path opened up by Claude Shannon in 1948. Andmore important still was the role that he played in defining, exemplifying, and promoting what has become the standard research paradigm in an even wider range of fields: the competitive evaluation of alternative algorithms based on precise quantitative criteria defined in advance, relative to a shared body of training and testing material. After receiving his Ph.D. from MIT in 1962, Fred taught at Cornell from 1962 to 1972, worked at IBM from 1972 to 1993, and taught at Johns Hopkins from 1993 to 2010. Fred’s many technical accomplishments during this long and productive career can be seen as episodes in two epic narratives, which, like the Iliad and the Odyssey, are related but have separate themes and story lines. The theme of the first epic is the return of Information Theory to center stage in speech and language processing; and the theme of the second epic is the development of a new relationship between science and engineering in speech recognition, computational linguistics, and artificial intelligence (AI) more generally. Fred gave us a vivid first-person narrative of the first of these epic adventures in his ACL Lifetime Achievement Award speech (Jelinek 2009). But missing from this recital is a picture of the dire state of the information-theoretic forces when Fred joined the fray by undertaking speech recognition research at IBM in 1972. To understand what the world was like then, we need to go back to the mid 1950s. Claude Shannon (1956, page 3) wrote: