FRANK FALLSIDE first became interested in speech-processing, a field in which he became a world authority, when he decided to use his technical skills to help deaf children to learn to speak clearly.
The ingenious method he devised for this was to invent a way of converting speech wave-forms, captured by a microphone, into visual representations of the vocal tract's behaviour. The speakers could then look at a screen and, under suitable guidance, learn to modify their speech articulation until it produced an intelligible and acceptable form, as indicated by the picture of the vocal tract, compared with that for correct speech. Thus sight was substituted for sound in overcoming the disability of deafness. The reward of so helping those with severe speech difficulties created a lifelong interest in speech-processing. His sudden death has deprived the world of one of its leading research workers in the speech-processing field.
Such a field of work requires a great depth of skills which take many years to acquire. After leaving George Heriot's School in Edinburgh, Fallside qualified as an electrical engineer at Edinburgh University, then completed his training by doing doctoral research at Swansea on the modelling and control of electrical machinery. This early work on non-linear dynamical system stability aroused in Fallside a deep and sustained interest in the use of computers for the analysis and synthesis of information-processing systems. He joined the Department of Engineering at Cambridge University in 1958, and became a Fellow of Trinity Hall in 1962. In the 1960s he worked on the analysis and control of power systems. From the early 1970s his work concentrated increasingly on the field of speech-processing, and he built up an enthusiastic and productive team in Cambridge with a prolific output of research publications. Their concentrated research effort eventually flowered into a broad attack on the use of information processing both to generate and to understand natural spoken language.
There are two fundamentally different ways of attacking such basic problems in any practical application of information-processing. One uses software - computer programs - to specify how the processing should be done; and the other uses hardware - the direct construction and use of appropriate forms of electrical circuit. A particular feature of much of Fallside's work was its emphasis on the use of artificial neural networks, which can be thought of as directly mimicking brain function. At the time when he started to put a heavy emphasis on the neural network approach, few people regarded it as likely to succeed. But this neural network approach is now recognised as of fundamental importance in a variety of important fields.
By the mid-1980s Fallside's group was firmly established as one of the world's leading research units in speech-processing, and as one having much wider interests, extending to vision processing and robotics. He created in the Engineering Department at Cambridge a multi-disciplinary postgraduate course in Computer Speech and Language Processing, and collaborated widely with cognate research groups in other parts of the university. Recently he had begun to develop an ambitious and highly promising new approach to language acquisition which sought to fuse the abilities to understand and to create speech. An active member of the international research community in his field, he was associated with the editing of many publications, including the journal Speech and Language Processing, which he founded in 1986.
It is increasingly difficult to combine high-quality research with the heavy burdens on the academic who is also committed to teaching and administration. Frank Fallside was a genial and even-tempered man who managed this difficult feat with conspicuous success. This he did by invariably taking an ultimately correct decision that someone, somewhere, somehow would eventually solve any problems generated by the creative use of space and other scarce resources. He was greatly liked by all who knew him well, and particularly by the very many research students he supervised, and by those research colleagues with whom he worked closely over more than two decades of sustained effort. His tragically early death has cut short an outstanding research career.