Answers are grounded in this paper and cite their sources.
This Professional Practice Report outlines a strategy for how to change the conclusion framework in forensic voice comparison from verbal posterior probabilities to verbal likelihood ratios at the Text, Speech and Audio Department of the German Bundeskriminalamt. While immediately intended for this purpose, the issues raised are of broader concern and should be of interest to other providers of forensic voice comparison services. The major issue addressed is how to combine results from different methods when these methods prefer different ways of stating the hypotheses in the likelihood ratio. As such combinations require the underlying hypotheses to be identical, suggestions are made about how the hypotheses can be expressed in a unitary fashion. The proposal arrived at is that the scope of the relevant population used in the hypotheses should specify language, gender, and broad age group (adulthood) but not language variety. It is proposed that methods that require reference to language variety rather than just language are treated with a nested form of the likelihood ratio in which language variety is analyzed as group-level evidence upon which the individual-level evidence is evaluated. It is also shown how group-level evidence can be evaluated without needing a separate definition of the relevant population from individual-level analysis. The second part of this report will address the distinctions common/specific source and score-/feature-based.
No takes yet. Share an insight, caveat, or question.
Michael E. Jessen (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: