{"url":"/sota/speech-recognition-on-switchboard-hub500","task":{"name":"Speech Recognition","url":"/task/speech-recognition","note":null},"dataset":{"name":"Switchboard + Hub500","url":null},"category":"Audio","categories":["Audio","Speech"],"category_note":null,"description":"**Speech Recognition** is the task of converting spoken language into text. It involves recognizing the words spoken in an audio recording and transcribing them into a written format. The goal is to accurately transcribe the speech in real-time or from recorded audio, taking into account factors such as accents, speaking speed, and background noise.\r\n\r\n<span style=\"color:grey; opacity: 0.6\">( Image credit: [SpecAugment](https://arxiv.org/pdf/1904.08779v2.pdf) )</span>","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["Percentage error"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Percentage error":"lower"}},"counts":{"rows":30,"rows_with_code":5,"rows_with_paper_page":17,"rows_dated":17,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"IBM (LSTM+Conformer encoder-decoder)","metrics":{"Percentage error":"4.3"},"uses_additional_data":false,"paper_date":"2021-05-03","paper":"/paper/on-the-limit-of-english-conversational-speech","paper_url":"https://arxiv.org/abs/2105.00982v1","paper_title":"On the limit of English conversational speech recognition","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"IBM (LSTM encoder-decoder)","metrics":{"Percentage error":"4.7"},"uses_additional_data":false,"paper_date":"2020-01-20","paper":"/paper/single-headed-attention-based-sequence-to","paper_url":"https://arxiv.org/abs/2001.07263v3","paper_title":"Single headed attention based sequence-to-sequence model for state-of-the-art results on Switchboard","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"ResNet + BiLSTMs acoustic model","metrics":{"Percentage error":"5.5"},"uses_additional_data":false,"paper_date":"2017-03-06","paper":"/paper/english-conversational-telephone-speech","paper_url":"http://arxiv.org/abs/1703.02136v1","paper_title":"English Conversational Telephone Speech Recognition by Humans and Machines","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"Microsoft 2016b","metrics":{"Percentage error":"5.8"},"uses_additional_data":false,"paper_date":"2016-10-17","paper":"/paper/achieving-human-parity-in-conversational","paper_url":"http://arxiv.org/abs/1610.05256v2","paper_title":"Achieving Human Parity in Conversational Speech Recognition","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":5,"model":"Microsoft 2016","metrics":{"Percentage error":"6.2"},"uses_additional_data":false,"paper_date":"2016-09-12","paper":"/paper/the-microsoft-2016-conversational-speech","paper_url":"http://arxiv.org/abs/1609.03528v2","paper_title":"The Microsoft 2016 Conversational Speech Recognition System","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":6,"model":"VGG/Resnet/LACE/BiLSTM acoustic model trained on SWB+Fisher+CH, N-gram + RNNLM language model trained on Switchboard+Fisher+Gigaword+Broadcast","metrics":{"Percentage error":"6.3"},"uses_additional_data":false,"paper_date":"2016-09-12","paper":"/paper/the-microsoft-2016-conversational-speech","paper_url":"http://arxiv.org/abs/1609.03528v2","paper_title":"The Microsoft 2016 Conversational Speech Recognition System","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":7,"model":"RNN + VGG + LSTM acoustic model trained on SWB+Fisher+CH, N-gram + \"model M\" + NNLM language model","metrics":{"Percentage error":"6.6"},"uses_additional_data":false,"paper_date":"2016-04-27","paper":"/paper/the-ibm-2016-english-conversational-telephone","paper_url":"http://arxiv.org/abs/1604.08242v2","paper_title":"The IBM 2016 English Conversational Telephone Speech Recognition System","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":8,"model":"CNN-LSTM","metrics":{"Percentage error":"6.6"},"uses_additional_data":false,"paper_date":"2016-10-17","paper":"/paper/achieving-human-parity-in-conversational","paper_url":"http://arxiv.org/abs/1610.05256v2","paper_title":"Achieving Human Parity in Conversational Speech Recognition","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":9,"model":"IBM 2016","metrics":{"Percentage error":"6.9"},"uses_additional_data":false,"paper_date":"2016-04-27","paper":"/paper/the-ibm-2016-english-conversational-telephone","paper_url":"http://arxiv.org/abs/1604.08242v2","paper_title":"The IBM 2016 English Conversational Telephone Speech Recognition System","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":10,"model":"RNNLM","metrics":{"Percentage error":"6.9"},"uses_additional_data":false,"paper_date":"2016-09-12","paper":"/paper/the-microsoft-2016-conversational-speech","paper_url":"http://arxiv.org/abs/1609.03528v2","paper_title":"The Microsoft 2016 Conversational Speech Recognition System","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":11,"model":"IBM 2015","metrics":{"Percentage error":"8.0"},"uses_additional_data":false,"paper_date":"2015-05-21","paper":"/paper/the-ibm-2015-english-conversational-telephone","paper_url":"http://arxiv.org/abs/1505.05899v1","paper_title":"The IBM 2015 English Conversational Telephone Speech Recognition System","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":12,"model":"HMM-BLSTM trained with MMI + data augmentation (speed) + iVectors + 3 regularizations + Fisher","metrics":{"Percentage error":"8.5"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":13,"model":"HMM-TDNN trained with MMI + data augmentation (speed) + iVectors + 3 regularizations + Fisher (10% / 15.1% respectively trained on SWBD only)","metrics":{"Percentage error":"9.2"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":14,"model":"CNN on MFSC/fbanks + 1 non-conv layer for FMLLR/I-Vectors concatenated in a DNN","metrics":{"Percentage error":"10.4"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":15,"model":"HMM-TDNN + iVectors","metrics":{"Percentage error":"11"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":16,"model":"CNN","metrics":{"Percentage error":"11.5"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":17,"model":"Deep CNN (10 conv, 4 FC layers), multi-scale feature maps","metrics":{"Percentage error":"12.2"},"uses_additional_data":false,"paper_date":"2015-09-29","paper":"/paper/very-deep-multilingual-convolutional-neural","paper_url":"http://arxiv.org/abs/1509.08967v2","paper_title":"Very Deep Multilingual Convolutional Neural Networks for LVCSR","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":18,"model":"HMM-DNN +sMBR","metrics":{"Percentage error":"12.6"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":19,"model":"DNN sMBR","metrics":{"Percentage error":"12.6"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":20,"model":"Deep Speech + FSH","metrics":{"Percentage error":"12.6"},"uses_additional_data":false,"paper_date":"2014-12-17","paper":"/paper/deep-speech-scaling-up-end-to-end-speech","paper_url":"http://arxiv.org/abs/1412.5567v2","paper_title":"Deep Speech: Scaling up end-to-end speech recognition","code":"https://github.com/mozilla/STT","n_code_links":24,"syntology":{"n_ran":9,"n_unverified":0,"n_samples":9,"n_pointer_only_licence":8}},{"rank_in_archive_order":21,"model":"CNN + Bi-RNN + CTC (speech to letters), 25.9% WER if trainedonlyon SWB","metrics":{"Percentage error":"12.6"},"uses_additional_data":false,"paper_date":"2014-12-17","paper":"/paper/deep-speech-scaling-up-end-to-end-speech","paper_url":"http://arxiv.org/abs/1412.5567v2","paper_title":"Deep Speech: Scaling up end-to-end speech recognition","code":"https://github.com/mozilla/STT","n_code_links":24,"syntology":{"n_ran":9,"n_unverified":0,"n_samples":9,"n_pointer_only_licence":8}},{"rank_in_archive_order":22,"model":"DNN MMI","metrics":{"Percentage error":"12.9"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":23,"model":"DNN MPE","metrics":{"Percentage error":"12.9"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":24,"model":"DNN BMMI","metrics":{"Percentage error":"12.9"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":25,"model":"HMM-TDNN + pNorm + speed up/down speech","metrics":{"Percentage error":"12.9"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":26,"model":"DNN + Dropout","metrics":{"Percentage error":"15"},"uses_additional_data":false,"paper_date":"2014-06-30","paper":"/paper/building-dnn-acoustic-models-for-large","paper_url":"http://arxiv.org/abs/1406.7806v2","paper_title":"Building DNN Acoustic Models for Large Vocabulary Speech Recognition","code":"https://github.com/pannous/caffe-speech-recognition","n_code_links":1,"syntology":null},{"rank_in_archive_order":27,"model":"DNN","metrics":{"Percentage error":"16"},"uses_additional_data":false,"paper_date":"2014-06-30","paper":"/paper/building-dnn-acoustic-models-for-large","paper_url":"http://arxiv.org/abs/1406.7806v2","paper_title":"Building DNN Acoustic Models for Large Vocabulary Speech Recognition","code":"https://github.com/pannous/caffe-speech-recognition","n_code_links":1,"syntology":null},{"rank_in_archive_order":28,"model":"CD-DNN","metrics":{"Percentage error":"16.1"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":29,"model":"DNN-HMM","metrics":{"Percentage error":"18.5"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":30,"model":"Deep Speech","metrics":{"Percentage error":"20"},"uses_additional_data":false,"paper_date":"2014-12-17","paper":"/paper/deep-speech-scaling-up-end-to-end-speech","paper_url":"http://arxiv.org/abs/1412.5567v2","paper_title":"Deep Speech: Scaling up end-to-end speech recognition","code":"https://github.com/mozilla/STT","n_code_links":24,"syntology":{"n_ran":9,"n_unverified":0,"n_samples":9,"n_pointer_only_licence":8}}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,885 of the 9,623 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9623,"papers_checked":6885,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2737},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-25T09:33:49+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":3,"rows_with_any_sample_ran":3,"distinct_papers_with_graph_line":1,"distinct_papers_with_any_sample_ran":1,"samples_over_distinct_papers":{"n_ran":9,"n_unverified":0,"n_samples":9,"n_pointer_only_licence":8,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":27,"n_unverified":0,"n_samples":27,"n_pointer_only_licence":24,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}