Cosine-based subsequence DTW over XLSR-53 embeddings retrieves spoken query words without transcription despite the embeddings' strong anisotropy.
This suggests that these representations capture high-level phonetic structures that generalize well across speakers, making them ro- bust for computational documentary linguistics
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.SD 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Assessing the Impact of Anisotropy in Neural Representations of Speech: A Case Study on Keyword Spotting
Cosine-based subsequence DTW over XLSR-53 embeddings retrieves spoken query words without transcription despite the embeddings' strong anisotropy.