Abstract:

This paper makes a case for forensic speaker-identification, but only if the practitioner is properly trained and carries out the task conscientiously. It could be argued (and has been argued) that it is unethical to engage in forensic speaker-identification until there is a well-established and fully automatic (i.e. machine-based) approach available: in other words phoneticians should not practise in this field at all until the subjective element of their task has been removed. The present contribution will focus on forensic speaker-profiling and identification. First, the specifics of the forensic task as opposed to the commercial speaker-identification (SI) task will be summarized, followed by a brief outline of the methods currently employed by forensic phoneticians. The applicability of automatic SI procedures will then be examined. It transpires that only in a small proportion of forensic cases does the material which is available from either the plaintiff or the investigating agency lend itself to the application of automatic methods. Therefore, it might seem unethical to apply these methods uncritically. However, in the vast majority of cases, other non-automatic methods have to be pursued. It is contended that the forensic phonetician has a moral obligation to aid the course of justice within the limitations which are imposed by the quantity and quality of the speech samples in question.

Author: Angelika Braun, Hermann J. Künzel