DI MASSIMO GRASSI, MARIA GRAZIA VACCARO, GIUSEPPE SARTORI
ABSTRACT – Audio recordings – such as environmental wiretaps – are often
used as evidence to support accusatory or defensive arguments.
Normally, these recordings are entrusted to a technical consultant or
expert who has the task of transcribing the content. Subsequently, the
debate takes place starting from and based on the transcribed text. This
passage hides difficulties which can become a serious problem if the
recordings contain incomprehensible material. We will propose a method
to quantify how intelligible the content of a given recording is.
Furthermore, we will propose a method for producing a transcription that
is maximally shareable. The contents and methods presented here concern
in particular the transcription of degraded or unintelligible audio, either
because it was recorded with poor quality equipment or techniques, or
because the recording conditions were not optimal. The text is divided
into three parts. In the first we will highlight how understanding the
content of unintelligible recordings is an extremely influential process. In
the second, we will propose the method for the numerical estimation of
intelligibility and the one for the reconstruction of a transcription that can
be maximally shared. In the third, both methods will be applied to a
recently debated murder case.
KEYWORDS – wiretapping – registration – intelligible – support accusatory – support defensive.