Layer-wise Probing of wav2vec 2.0 and Whisper for Consonant Cluster Reduction in African American English

Hamid, Kevin · 2026

Read Layer-wise Probing of wav2vec 2.0 and Whisper for Consonant Cluster Reduction in African American English by Hamid, Kevin (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-23948.txt.

Description

Mojarad and Tang probe the inner layers of two speech recognition models to see how they encode a hallmark sound pattern of African American English, in a 2026 paper on speech science. Consonant cluster reduction simplifies final clusters, so a word like test may surface with its last stop weakened. The pattern is rule governed, yet recognizers err far more on this dialect. The authors run layer wise probes on wav2vec 2.0 and Whisper, testing whether each layer can tell reduced from full forms and recover the underlying cluster.

Each model separates reduced forms from canonical ones with high accuracy at its best layers. Even when a stop is barely pronounced, the representations retain cues to which stop belongs there. Reduction is thus encoded as structured, gradient variation, not as plain deletion. Modern speech encoders capture the dialect's phonology in an organized way.

Catalog record

CREATORHamid, Kevin
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.23948

Browse the full library · About the Tower of Babel