English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

We present our analysis of a significant data artifact in the official2019/2021 ASVspoof Challenge Dataset. We identify an uneven distribution ofsilence duration in the training and test splits, which tends to correlate withthe target prediction label. Bonafide instances tend to have significantlylonger leading and trailing silences than spoofed instances. In this paper, weexplore this phenomenon and its impact in depth. We compare several types ofmodels trained on a) only the duration of the leading silence and b) only onthe duration of leading and trailing silence. Results show that models trainedon only the duration of the leading silence perform particularly well, andachieve up to 85% percent accuracy and an equal error rate (EER) of 15.1%. Atthe same time, we observe that trimming silence during pre-processing and thentraining established antispoofing models using signal-based features leads tocomparatively worse performance. In that case, EER increases from 3.6% (withsilence) to 15.5% (trimmed silence). Our findings suggest that previous workmay, in part, have inadvertently learned thespoof/bonafide distinction byrelying on the duration of silence as it appears in the official challengedataset. We discuss the potential consequences that this has for interpretingsystem scores in the challenge and discuss how the ASV community may furtherconsider this issue.

Speech is Silver, Silence is Golden: What do ASVspoof-trained Models Really Learn?