Issues of bilingualism in likelihood ratio-based forensic voice comparison
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Situated at the intersection of forensic speech science and bilingualism, this thesis focuses on the issues of language and language mismatch in forensic voice comparison (FVC) and examines their effects on features commonly used in FVC within the framework of likelihood ratios (LRs). To this end, two experiments are presented which explore (1) the performance of the alveolar fricative /s/, long-term formant distributions (LTFDs) and automatic speaker recognition (ASR) software as speaker discriminants in same-language comparisons in Canadian English and French, and (2) the performance of the features above in cross-language comparisons, following a cross-linguistic acoustic analysis of the linguistic-phonetic features. \n \nAlthough /s/ showed stronger language-independence acoustically than LTFDs, results from Experiment 1 show that /s/ performed more strongly as a speaker discriminant in French than in English, whereas the performance of LTFDs and ASR in the two languages was similar. Results from Experiment 2 show poorer performance across all features to varying extents in cross-language comparisons, which was exacerbated when appropriate reference data matching the language conditions of the case were not used. Individual-level analysis further reveals a complex mapping between acoustic and individual performance in cross-language comparisons. In particular, speakers for whom LTFDs provided the strongest discriminatory performance did not necessarily show the lowest within-speaker variation. \n \nOverall, findings from the current study contribute to our understanding of cross-language comparisons, and more generally to the area of forensic speech science, by demonstrating quantitatively the impact of language mismatch on the discriminatory potential of different linguistic-phonetic and acoustic features within the numerical LR framework, as well as the significance of case-appropriate reference data in such cases. They also demonstrate the diagnostic value of individual-level analysis in system testing and indicate the need for a more nuanced conception of within- and between-speaker variability for FVC.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.003 | 0.000 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it