harvard-edge/multilingual_kws

Issues

• How to train it for more than one target_keyword?

  #36 opened 9 months ago by twshen2000
  3

• dataset of multilingual_context_73_0.8011

  #42 opened a year ago by farzadhallaji
  1

• Where can I find the used keywords (total 760) and splits from the paper?

  #41 opened 2 years ago by V0XNIHILI
  3

• Reproducing paper results

  #39 opened 2 years ago by sathibault
  0

• OperatorNotAllowedInGraphError during Transfer Learning

  #38 opened 2 years ago by ccioflan
  0

• ERROR: Cannot find key when Running docker

  #37 opened 2 years ago by manarsaaldossari
  0

• Using multilingual_kws with microphone streaming

  #34 opened 3 years ago by wesbz
  5

• some empty directories in MSWC? or the 16KHz reencode?

  #35 opened 3 years ago by mmaz
  0

• add version info to each tarball

  #31 opened 3 years ago by mmaz
  0

• GCS transfer

  #28 opened 3 years ago by mmaz
  1

• is Mozilla SWTS in the dataset?

  #25 opened 3 years ago by mmaz
  2

• Re-encode mp3/opus clips to exactly 1s

  #8 opened 3 years ago by mmaz
  0

• First time user

  #3 opened 3 years ago by El-Yazid
  4

• UMAP visualization transitive dependency on old numpy

  #33 opened 3 years ago by mmaz
  0

• words with apostrophes are not correctly being extracted

  #32 opened 3 years ago by mmaz
  0

• Filter out NaNs from Common Voice tsvs, distinguish between intentional "nan" in language vocabulary

  #9 opened 3 years ago by mmaz
  0

• found duplicates at __2 and __3 etc

  #20 opened 3 years ago by mmaz
  3

• Lithuanian clips not validated

  #11 opened 3 years ago by Sharad24
  1

• Arabic word formatting

  #29 opened 3 years ago by mmaz
  1

• expand and validate text normalization/cleaning filters

  #27 opened 3 years ago by mmaz
  1

• TFDS api

  #23 opened 3 years ago by mmaz
  2

• words greater than 2^3 are probably > 1s

  #22 opened 3 years ago by mmaz
  1

• generate a sibling dataset with speech context

  #15 opened 3 years ago by mmaz
  1

• Some percentage of wavs (~3%) are below 1s according to soxi, others can't be opened

  #10 opened 3 years ago by mmaz
  1

• Add TFDS integration/flow

  #24 opened 3 years ago by colbybanbury
  0

• Bug: Some languages (basque, polish) seem to have a higher total length duration than in original common Voice

  #13 opened 3 years ago by Sharad24
  0

• re-run few shot experiments with same split of unknown/silence as the DSCNN tests

  #19 opened 3 years ago by mmaz
  0

• rerun DSCNN tests with fixed AudioSeed data for unknown sampling

  #17 opened 3 years ago by mmaz
  1

• verify 1-1 match between final audio files and splits

  #14 opened 3 years ago by mmaz
  0

• Re-creating alignments for Common Voice 7

  #12 opened 3 years ago by Sharad24
  0

• check for 16KHz in AudioDataset

  #7 opened 3 years ago by mmaz
  0

• is there the inference.py？

  #4 opened 3 years ago by huacilang
  1

• word counts

  #1 opened 3 years ago by mmaz
  2