Raj Reddy

Overview

Raj Reddy is a computer scientist who spent most of his working life at Carnegie Mellon University (CMU), where he built speech recognition systems and founded and led large research organizations. His name is attached both to a line of speech recognition programs and to the institutions that trained a generation of researchers in artificial intelligence and robotics. In 1994 he was co-recipient of the ACM Turing Award, jointly with Ed Feigenbaum ^[1].

Education and early career

Dabbala Rajagopal (Raj) Reddy was born on June 13, 1937, in Katoor, Andhra Pradesh, India ^[1]. His first degree was not in computing: he received a B.S. in civil engineering from Guindy College of Engineering in Madras — now Anna University, Chennai — in 1958, followed in 1960 by an MTech from the University of New South Wales in Sydney, Australia ^[1]. From 1960 to 1963 he worked in Australia as an Applied Science Representative for IBM ^[1].

He then moved to the United States to study at Stanford University, where he completed his PhD in 1966 under the supervision of John McCarthy, with a dissertation on speech recognition; it was the first PhD granted by Stanford's newly formed Department of Computer Science ^[1].

Speech recognition

Reddy's central technical contribution is a series of speech recognition systems. He developed Hearsay I, the first system capable of continuous speech recognition, and with his students built the systems that followed it, including Hearsay II, Harpy and Dragon ^[1].

Continuous speech recognition is the harder version of the problem: the machine has to follow ordinary speech, in which words run into one another, rather than a short sequence of words spoken one at a time with clear pauses between them. That calls for pulling together different kinds of evidence — how speech sounds are produced and perceived, what sounds and words a language contains, and which sequences of words are permissible — and getting them to agree on a single answer.

Much of the way modern commercial speech recognition is put together traces back to this work: in these systems Reddy and his students developed most of the ideas underlying modern commercial speech recognition technology, including the "blackboard model" for coordinating multiple knowledge sources ^[1].

A blackboard model is a way of organizing a system in which no single program holds the whole answer. Independent knowledge sources — each an expert on one kind of evidence — post partial conclusions to a shared workspace, the "blackboard", and a control part decides which expert should look at the current state next. Each new posting can confirm, revise or extend what is already there, so the interpretation of an utterance is built up in steps rather than in one pass. The arrangement makes it possible to add or change one source of knowledge without rewriting the others.

Institutional roles

Reddy's academic career moved between Stanford and Carnegie Mellon. He was Assistant Professor of Computer Science at Stanford from 1966 to 1969, then Associate Professor at Carnegie Mellon from 1969, and University Professor of Computer Science there from 1984 ^[1].

At Carnegie Mellon he took on organizing work on a large scale. He was Founding Director of the Robotics Institute from 1980 to 1992, and Dean of the School of Computer Science from 1991 to 1999, so the later years of his directorship of the institute overlapped with the start of his deanship ^[1].

He was also active in the professional community: he served as President of the American Association for Artificial Intelligence from 1987 to 1989 ^[1].

Recognition

Reddy's honours span several countries. In 1984 he received the Legion d'Honneur, presented by President Mitterrand of France, and was elected to the National Academy of Engineering ^[1]. In 1994 he was co-recipient of the ACM Turing Award jointly with Ed Feigenbaum ^[1]. In 2001 he received the Padma Bhushan Award, presented by the President of India, and in 2005 the Honda Prize ^[1].

Place in the field

Two strands run through Reddy's career. One is technical: the Hearsay and later systems, and the ideas about coordinating separate sources of knowledge that were carried into commercial speech technology ^[1]. The other is organizational. The Robotics Institute and the School of Computer Science at Carnegie Mellon, both of which he led from their early years, gave a home to work that mixes computing with the physical world, and his students took part in building the speech systems alongside him ^[1]. That combination — a researcher who also spends decades constructing the institutions in which research happens — is what the Turing Award, the national academy membership and the other honours above record ^[1].