Evaluation of feature extraction methods for query-by-example spoken term detection with low resource languages
Published September 13, 2021 | Version v2
Dataset Open

Features for QbE-STD, wav2vec 2.0 XLSR-53 (Archive I)

Authors/Creators

  • 1. Stanford University
  • 2. University of Groningen

Description

Features extracted from audio using wav2vec 2.0 feature extractor with the XLSR-53 checkpoint for eng-mav, gbb-lg, wbp-jk, and wrl-mb datasets.

 

Files

eng-mav.zip

Files (45.1 GB)

Name Size
md5:ade1227e46484dbc3c114ba8f6745f35
24.6 GB Preview Download
md5:5027c31b2ce7bb9daaf8b9b7e10290eb
14.9 GB Preview Download
md5:4defd74e392e705b27fabcdff316b1ad
2.0 GB Preview Download
md5:11fff36bd2f639f6b75b1a4516256cbe
3.6 GB Preview Download