[
  {
    "Word Error Rate (WER)": 5,
    "code_links": [
      {
        "title": "rwth-i6/returnn",
        "url": "https://github.com/rwth-i6/returnn"
      },
      {
        "title": "rwth-i6/returnn-experiments",
        "url": "https://github.com/rwth-i6/returnn-experiments/tree/master/2019-librispeech-system"
      }
    ],
    "date": "2019-05-08",
    "date2": 20190508,
    "model": "Hybrid model with Transformer rescoring",
    "paper": {
      "title": "RWTH ASR Systems for LibriSpeech: Hybrid vs Attention -- w/o Data Augmentation",
      "url": "https://cknow.io/lib/3c4bbcadf9194f9e"
    },
    "paper_data_uoa": "3c4bbcadf9194f9e"
  },
  {
    "Word Error Rate (WER)": 5.7,
    "code_links": [
      {
        "title": "rwth-i6/returnn",
        "url": "https://github.com/rwth-i6/returnn"
      },
      {
        "title": "rwth-i6/returnn-experiments",
        "url": "https://github.com/rwth-i6/returnn-experiments/tree/master/2019-librispeech-system"
      }
    ],
    "date": "2019-05-08",
    "date2": 20190508,
    "model": "6 layer BLSTM with LSTM transformer rescoring",
    "paper": {
      "title": "RWTH ASR Systems for LibriSpeech: Hybrid vs Attention -- w/o Data Augmentation",
      "url": "https://cknow.io/lib/3c4bbcadf9194f9e"
    },
    "paper_data_uoa": "3c4bbcadf9194f9e"
  },
  {
    "Word Error Rate (WER)": 5.7,
    "code_links": [
      {
        "title": "espnet/espnet",
        "url": "https://github.com/espnet/espnet"
      }
    ],
    "date": "2019-09-13",
    "date2": 20190913,
    "model": "Transformer",
    "paper": {
      "title": "A Comparative Study on Transformer vs RNN in Speech Applications",
      "url": "https://cknow.io/lib/d68911301970abbe"
    },
    "paper_data_uoa": "d68911301970abbe"
  },
  {
    "Word Error Rate (WER)": 5.8,
    "code_links": [
      {
        "title": "mozilla/DeepSpeech",
        "url": "https://github.com/mozilla/DeepSpeech"
      },
      {
        "title": "shelling203/SpecAugment",
        "url": "https://github.com/shelling203/SpecAugment"
      },
      {
        "title": "DemisEom/SpecAugment",
        "url": "https://github.com/DemisEom/SpecAugment"
      },
      {
        "title": "lRomul/argus-freesound",
        "url": "https://github.com/lRomul/argus-freesound"
      },
      {
        "title": "ebouteillon/freesound-audio-tagging-2019",
        "url": "https://github.com/ebouteillon/freesound-audio-tagging-2019"
      },
      {
        "title": "sh951011/Korean-Speech-Recognition",
        "url": "https://github.com/sh951011/Korean-Speech-Recognition/blob/master/package/feature.py"
      },
      {
        "title": "sh951011/Korean-Speech-Recognition",
        "url": "https://github.com/sh951011/Korean-Speech-Recognition"
      },
      {
        "title": "KimJeongSun/SpecAugment_numpy_scipy",
        "url": "https://github.com/KimJeongSun/SpecAugment_numpy_scipy"
      },
      {
        "title": "hgstudent/las",
        "url": "https://github.com/hgstudent/las"
      },
      {
        "title": "viig99/mixmatch-freesound",
        "url": "https://github.com/viig99/mixmatch-freesound"
      },
      {
        "title": "knlee-voice/PaperNotes",
        "url": "https://github.com/knlee-voice/PaperNotes"
      }
    ],
    "date": "2019-04-18",
    "date2": 20190418,
    "model": "LAS + SpecAugment",
    "paper": {
      "title": "SpecAugment: A Simple Data Augmentation Method for Automatic Speech Recognition",
      "url": "https://cknow.io/lib/15408ea14d4c82f9"
    },
    "paper_data_uoa": "15408ea14d4c82f9"
  },
  {
    "Word Error Rate (WER)": 5.8,
    "code_links": [
      {
        "title": "s-omranpour/ConvolutionalSpeechRecognition",
        "url": "https://github.com/s-omranpour/ConvolutionalSpeechRecognition"
      }
    ],
    "date": "2019-10-01",
    "date2": 20191001,
    "model": "Multi-Stream Self-Attention With Dilated 1D Convolutions",
    "paper": {
      "title": "State-of-the-Art Speech Recognition Using Multi-Stream Self-Attention With Dilated 1D Convolutions",
      "url": "https://cknow.io/lib/35a2912594df0b5b"
    },
    "paper_data_uoa": "35a2912594df0b5b"
  },
  {
    "Word Error Rate (WER)": 7.63,
    "code_links": [],
    "date": "2018-04-15",
    "date2": 20180415,
    "model": "tdnn + chain + rnnlm rescoring",
    "paper": {
      "title": "Neural Network Language Modeling with Letter-based Features and Importance Sampling",
      "url": "https://cknow.io/lib/5449c55a2744f069"
    },
    "paper_data_uoa": "5449c55a2744f069"
  },
  {
    "Word Error Rate (WER)": 8.79,
    "code_links": [
      {
        "title": "NVIDIA/OpenSeq2Seq",
        "url": "https://github.com/NVIDIA/OpenSeq2Seq"
      }
    ],
    "date": "2019-04-05",
    "date2": 20190405,
    "model": "deep 1d convs + ctc + external lm rescoring",
    "paper": {
      "title": "Jasper: An End-to-End Convolutional Neural Acoustic Model",
      "url": "https://cknow.io/lib/d52115ec6c741b09"
    },
    "paper_data_uoa": "d52115ec6c741b09"
  },
  {
    "Word Error Rate (WER)": 10.47,
    "code_links": [],
    "date": "2018-12-17",
    "date2": 20181217,
    "model": "Convolutional Speech Recognition",
    "paper": {
      "title": "Fully Convolutional Speech Recognition",
      "url": "https://cknow.io/lib/d1db60cb34872a82"
    },
    "paper_data_uoa": "d1db60cb34872a82"
  },
  {
    "Word Error Rate (WER)": 13.25,
    "code_links": [
      {
        "title": "tensorflow/models",
        "url": "https://github.com/tensorflow/models/tree/master/research/deep_speech"
      },
      {
        "title": "PaddlePaddle/models",
        "url": "https://github.com/PaddlePaddle/models"
      },
      {
        "title": "baidu-research/warp-ctc",
        "url": "https://github.com/baidu-research/warp-ctc"
      },
      {
        "title": "SeanNaren/deepspeech.pytorch",
        "url": "https://github.com/SeanNaren/deepspeech.pytorch"
      },
      {
        "title": "DeepMark/deepmark",
        "url": "https://github.com/DeepMark/deepmark"
      },
      {
        "title": "SeanNaren/deepspeech.torch",
        "url": "https://github.com/SeanNaren/deepspeech.torch"
      },
      {
        "title": "robmsmt/KerasDeepSpeech",
        "url": "https://github.com/robmsmt/KerasDeepSpeech"
      },
      {
        "title": "myrtleSoftware/deepspeech",
        "url": "https://github.com/myrtleSoftware/deepspeech"
      },
      {
        "title": "noahchalifour/deepspeech2-tensorflow",
        "url": "https://github.com/noahchalifour/deepspeech2-tensorflow"
      },
      {
        "title": "UnofficialJuliaMirror/DeepMark-deepmark",
        "url": "https://github.com/UnofficialJuliaMirror/DeepMark-deepmark"
      },
      {
        "title": "UnofficialJuliaMirrorSnapshots/DeepMark-deepmark",
        "url": "https://github.com/UnofficialJuliaMirrorSnapshots/DeepMark-deepmark"
      },
      {
        "title": "huylenguyen806/vnasr",
        "url": "https://github.com/huylenguyen806/vnasr"
      }
    ],
    "date": "2015-12-08",
    "date2": 20151208,
    "model": "Deep Speech 2",
    "paper": {
      "title": "Deep Speech 2: End-to-End Speech Recognition in English and Mandarin",
      "url": "https://cknow.io/lib/ae2bdef963023028"
    },
    "paper_data_uoa": "ae2bdef963023028"
  },
  {
    "Word Error Rate (WER)": 15.28,
    "code_links": [
      {
        "title": "facebookresearch/wav2letter",
        "url": "https://github.com/facebookresearch/wav2letter"
      }
    ],
    "date": "2020-02-24",
    "date2": 20200224,
    "model": "Local Prior Matching (Large Model, ConvLM LM)",
    "paper": {
      "title": "Semi-Supervised Speech Recognition via Local Prior Matching",
      "url": "https://cknow.io/lib/85675128034707b7"
    },
    "paper_data_uoa": "85675128034707b7"
  },
  {
    "Word Error Rate (WER)": 16.5,
    "code_links": [
      {
        "title": "snipsco/snips-nlu",
        "url": "https://github.com/snipsco/snips-nlu"
      },
      {
        "title": "snipsco/nlu-benchmark",
        "url": "https://github.com/snipsco/nlu-benchmark"
      },
      {
        "title": "MiuLab/SlotGated-SLU",
        "url": "https://github.com/MiuLab/SlotGated-SLU"
      }
    ],
    "date": "2018-05-25",
    "date2": 20180525,
    "model": "Snips",
    "paper": {
      "title": "Snips Voice Platform: an embedded Spoken Language Understanding system for private-by-design voice interfaces",
      "url": "https://cknow.io/lib/4466ad0dd77ea158"
    },
    "paper_data_uoa": "4466ad0dd77ea158"
  },
  {
    "Word Error Rate (WER)": 20.84,
    "code_links": [
      {
        "title": "facebookresearch/wav2letter",
        "url": "https://github.com/facebookresearch/wav2letter"
      }
    ],
    "date": "2020-02-24",
    "date2": 20200224,
    "model": "Local Prior Matching (Large Model)",
    "paper": {
      "title": "Semi-Supervised Speech Recognition via Local Prior Matching",
      "url": "https://cknow.io/lib/85675128034707b7"
    },
    "paper_data_uoa": "85675128034707b7"
  }
]