# Speech Recognition on Wall Street Journal open vocabulary (dev93) Source: [Speech Recognition on Wall Street Journal open vocabulary (dev93)]() Collection index: [llms.txt]() Data updated: 2026-02-26T06:08:24.477Z Evaluation of Speech Recognition performance on the dev93 split of the Wall Street Journal open vocabulary corpus, typically measured using Word Error Rate (WER). A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Speech Recognition | | Dataset (including any stated split) | Wall Street Journal open vocabulary (dev93) | | Metric | WER | | Unit | PERCENT | | Direction | Lower is better | | Results loaded | 28 of 28 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [WER](): Lower is better; unit: PERCENT. - [LER](): Lower is better; unit: PERCENT. ## Recorded results | Method | WER | Result date | Report paper | | --- | --- | --- | --- | | Deep Speech 2 | 4.4 | Not recorded | 68fa66831e209131cea10cc8 | | Deep Speech 2 | 4.42 | Not recorded | 68fa66551e209131cea0f8cc | | Supervised transfer-learning | 4.99 | Not recorded | 68fa66551e209131cea0f8cc | | DeepSpeech 2 | 5 | Not recorded | 68fa65a21e209131cea0b741 | | wav2vec large | 5.1 | Not recorded | 68fa66551e209131cea0f8cc | | wav2vec large | 5.16 | Not recorded | 68fa66551e209131cea0f8cc | | wav2vec | 5.31 | Not recorded | 68fa66551e209131cea0f8cc | | wav2vec | 5.39 | Not recorded | 68fa66551e209131cea0f8cc | | Lattice-free MMI | 5.66 | Not recorded | 68fa66551e209131cea0f8cc | | ESPRESSO | 5.9 | Not recorded | 68fa66831e209131cea10cc8 | | ESPRESSO | 5.9 | Not recorded | 68fa66831e209131cea10cc8 | | Log-mel filterbanks (Baseline) | 6.27 | Not recorded | 68fa66551e209131cea0f8cc | | Likhomanenko et al. | 6.4 | Not recorded | 68fa66831e209131cea10cc8 | | CNN-DNN-BLSTM-HMM | 6.6 | Not recorded | 68fa65a21e209131cea0b741 | | Log-mel filterbanks (Baseline) | 6.67 | Not recorded | 68fa66551e209131cea0f8cc | | Learnable front-end | 6.8 | Not recorded | 68fa65a21e209131cea0b741 | | Trainable frontend | 6.8 | Not recorded | 68fa66551e209131cea0f8cc | | Zeghidour et al. | 6.8 | Not recorded | 68fa66831e209131cea10cc8 | | Learnable front-end | 6.9 | Not recorded | 68fa65a21e209131cea0b741 | | wav2vec large | 6.96 | Not recorded | 68fa66551e209131cea0f8cc | | wav2vec | 7.4 | Not recorded | 68fa66551e209131cea0f8cc | | ESPRESSO | 7.4 | Not recorded | 68fa66831e209131cea10cc8 | | Mel-filterbanks | 7.5 | Not recorded | 68fa65a21e209131cea0b741 | | wav2vec | 7.59 | Not recorded | 68fa66551e209131cea0f8cc | | Log-mel filterbanks (Baseline) | 8.57 | Not recorded | 68fa66551e209131cea0f8cc | | wav2vec | 9.11 | Not recorded | 68fa66551e209131cea0f8cc | | Mel-filterbanks | 9.5 | Not recorded | 68fa65a21e209131cea0b741 | | ESPRESSO | 14.8 | Not recorded | 68fa66831e209131cea10cc8 | ### Deep Speech 2 WER: 4.4. Result ID: 698cf67e8b517eb4331e04aa. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Deep Speech 2 WER: 4.42. Result ID: 698ce9a18b517eb4331d9f21. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"labeled speech":"12K h"}. This record does not assert independent reproduction. ### Supervised transfer-learning WER: 4.99. Result ID: 698ce9a18b517eb4331d9f24. [Reporting paper]() Recorded evaluation configuration: {"phoneme-based":true}. This record does not assert independent reproduction. ### DeepSpeech 2 WER: 5. Result ID: 698cbcbe8b517eb4331c61f1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training hours":"12k","Acoustic Model (AM)":"true","LM":"common crawl"}. This record does not assert independent reproduction. ### wav2vec large WER: 5.1. Result ID: 698ce9a18b517eb4331d9f2f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Language Model":"CHAR CONVLM","Pre-training Data":"Librispeech","Pre-training Hours":"960h"}. This record does not assert independent reproduction. ### wav2vec large WER: 5.16. Result ID: 698ce9a18b517eb4331d9f2c. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Language Model":"WORD CONVLM","Pre-training Data":"Librispeech","Pre-training Hours":"960h"}. This record does not assert independent reproduction. ### wav2vec WER: 5.31. Result ID: 698ce9a18b517eb4331d9f2e. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Language Model":"CHAR CONVLM","Pre-training Data":"Librispeech","Pre-training Hours":"960h"}. This record does not assert independent reproduction. ### wav2vec WER: 5.39. Result ID: 698ce9a18b517eb4331d9f2b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Language Model":"WORD CONVLM","Pre-training Data":"Librispeech","Pre-training Hours":"960h"}. This record does not assert independent reproduction. ### Lattice-free MMI WER: 5.66. Result ID: 698ce9a18b517eb4331d9f23. [Reporting paper]() Recorded evaluation configuration: {"phoneme-based":true}. This record does not assert independent reproduction. ### ESPRESSO WER: 5.9. Result ID: 698cf67e8b517eb4331e04ad. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Architecture":"LSTM","Look-ahead Word LM":true,"Improved Coverage":true,"EOS Threshold":false,"Beam Size":50}. This record does not assert independent reproduction. ### ESPRESSO WER: 5.9. Result ID: 698cf67e8b517eb4331e04ae. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Architecture":"LSTM","Look-ahead Word LM":true,"Improved Coverage":true,"EOS Threshold":true,"Beam Size":50}. This record does not assert independent reproduction. ### Log-mel filterbanks (Baseline) WER: 6.27. Result ID: 698ce9a18b517eb4331d9f2a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Language Model":"WORD CONVLM"}. This record does not assert independent reproduction. ### Likhomanenko et al. WER: 6.4. Result ID: 698cf67e8b517eb4331e04a8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### CNN-DNN-BLSTM-HMM WER: 6.6. Result ID: 698cbcbe8b517eb4331c61f0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Speaker adaptation":"true","Acoustic states":"3k"}. This record does not assert independent reproduction. ### Log-mel filterbanks (Baseline) WER: 6.67. Result ID: 698ce9a18b517eb4331d9f2d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Language Model":"CHAR CONVLM"}. This record does not assert independent reproduction. ### Learnable front-end WER: 6.8. Result ID: 698cbcbe8b517eb4331c61f5. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Filters":"80","Language Model":"ConvLM"}. This record does not assert independent reproduction. ### Trainable frontend WER: 6.8. Result ID: 698ce9a18b517eb4331d9f22. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Zeghidour et al. WER: 6.8. Result ID: 698cf67e8b517eb4331e04a9. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Learnable front-end WER: 6.9. Result ID: 698cbcbe8b517eb4331c61f4. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Filters":"40","Language Model":"ConvLM"}. This record does not assert independent reproduction. ### wav2vec large WER: 6.96. Result ID: 698ce9a18b517eb4331d9f29. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Language Model":"4-GRAM LM","Pre-training Data":"Librispeech","Pre-training Hours":"960h"}. This record does not assert independent reproduction. ### wav2vec WER: 7.4. Result ID: 698ce9a18b517eb4331d9f27. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Language Model":"4-GRAM LM","Pre-training Data":"Librispeech","Pre-training Hours":"960h"}. This record does not assert independent reproduction. ### ESPRESSO WER: 7.4. Result ID: 698cf67e8b517eb4331e04ac. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Architecture":"LSTM","Look-ahead Word LM":true,"Improved Coverage":false,"EOS Threshold":false,"Beam Size":50}. This record does not assert independent reproduction. ### Mel-filterbanks WER: 7.5. Result ID: 698cbcbe8b517eb4331c61f3. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Language Model":"ConvLM"}. This record does not assert independent reproduction. ### wav2vec WER: 7.59. Result ID: 698ce9a18b517eb4331d9f28. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Language Model":"4-GRAM LM","Pre-training Data":"Libri + WSJ","Pre-training Hours":"1,041 h"}. This record does not assert independent reproduction. ### Log-mel filterbanks (Baseline) WER: 8.57. Result ID: 698ce9a18b517eb4331d9f25. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Language Model":"4-GRAM LM"}. This record does not assert independent reproduction. ### wav2vec WER: 9.11. Result ID: 698ce9a18b517eb4331d9f26. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Language Model":"4-GRAM LM","Pre-training Data":"Librispeech","Pre-training Hours":"80h"}. This record does not assert independent reproduction. ### Mel-filterbanks WER: 9.5. Result ID: 698cbcbe8b517eb4331c61f2. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Language Model":"4-gram"}. This record does not assert independent reproduction. ### ESPRESSO WER: 14.8. Result ID: 698cf67e8b517eb4331e04ab. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Architecture":"LSTM","Look-ahead Word LM":false,"Improved Coverage":false,"EOS Threshold":false,"Beam Size":50}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.