Contemporary research and practice in speech science and technology are driven by the availability of large speech corpora. While audio databases exist for languages spoken mainly in America, Europe and Japan, there is currently no large auditory-visual database of spoken language, and certainly not one for Australian English. In this paper a recent ARC LIEF grant application to remedy this state of affairs is described. If funded (and by the time this paper is presented the outcome will be known) the project will involve 27 speech scientists from 12 institutions banded together via the Australasian Speech Science and Technology Association to record audiovisually, using a standard protocol and hardware set-up, speech in various contexts on three separate occasions. In total there will be 744 speakers of standard Australian English, 192 of ethnocultural variations, 96 of Aboriginal Australian English or creoles, and 96 with disordered speech; a total of 3384 recording sessions, the results of which will establish the Australian Speech Science Infrastructure (AASI). The ASSI would support a number of current projects on Australian English and facilitate many new research projects in cognitive science, phonetics, engineering, for example in areas such as audiovisual speech recognition, new avatar-based web technologies, and a range of security, speech and speaker recognition and forensic applications.
Authors: Denis Burnham
Event: SF08: Designing the Australian National Corpus Workshop