Spoken Language Technology, Speech Science and Human Communication

While homo sapiens is only one of many species that have gained an evolutionary advantage by communicating through sound waves, the complexity of human speech communication and the information density of human speech sounds distinguishes us clearly from other species. We are able to convolve - consciously and unconsciously - layers upon layers of information into a stream of sounds, from which those who are close to us will - again both consciously and unconsciously - elicit much of the information we encoded in the first place. This paper looks at human communication from the point of view of the spoken language engineer who will typically draw on a model that represents only one specific aspect of the human-to-human communication in order to address a particular communication problem; to enhance human-to-human communication in some way; to build a machine that produces sound streams recognisable by humans as "speech"; or to build a machine that can decode part of the information conveyed by human speech. Engineering tasks from the application areas of communication technology, speech and speaker recognition, speech synthesis and speech-technology tools for the handicapped are examined with respect to the specific models of human communication employed to solve those tasks and with respect to the interplay between the speech sciences and the engineering tasks at hand. The paper concludes with a brief exploration of how human communication itself has changed as a result of the advances of speech science and speech technology.