TasksSotADatasetsPapersMethodsSubmitAbout
Papers With Code 2

A community resource for machine learning research: papers, code, benchmarks, and state-of-the-art results.

Explore

Notable BenchmarksAll SotADatasetsPapersMethods

Community

Submit ResultsAbout

Data sourced from the PWC Archive (CC-BY-SA 4.0). Built by the community, for the community.

Models/SpeechStew (100M)

SpeechStew (100M)

Reported on 7 benchmarks across 1 task · 1 paper · 5 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Audio7 results

  • Speech RecognitiononSwitchboard CallHome
    Word Error Rate (WER)· 2021-04-05
    8.3
    SOTA
    SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural NetworkarXiv:2104.02133
  • Speech RecognitiononAMI IMH
    Word Error Rate (WER)· 2021-04-05
    9
    best: 7.8 (ConformerXXL-P + Downstream NST)
    SOTA
    SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural NetworkarXiv:2104.02133
  • Speech RecognitiononTedlium
    Word Error Rate (WER)· 2021-04-05
    5.3
    best: 0.29 (United-MedASR (764M))
    SOTA
    SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural NetworkarXiv:2104.02133
  • Speech RecognitiononSwitchboard SWBD
    Word Error Rate (WER)· 2021-04-05
    4.7
    SOTA
    SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural NetworkarXiv:2104.02133
  • Speech RecognitiononAMI SDM1
    Word Error Rate (WER)· 2021-04-05
    21.7
    best: 17.7 (ConformerXXL-P)
    SOTA
    SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural NetworkarXiv:2104.02133
  • Speech RecognitiononLibriSpeech test-clean
    Word Error Rate (WER)· 2021-04-05
    2
    best: 0.985 (United Med ASR)
    SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural NetworkarXiv:2104.02133
  • Speech RecognitiononLibriSpeech test-other
    Word Error Rate (WER)· 2021-04-05
    4
    best: 2.48 (SAMBA ASR)
    SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural NetworkarXiv:2104.02133