--- dataset_info: - config_name: cymen_arfor__lleisiau_arfor features: - name: sentence dtype: string - name: accent dtype: string - name: language dtype: string - name: id dtype: string - name: prediction dtype: string splits: - name: openai__whisper_large_v3__main num_bytes: 642060 num_examples: 3735 - name: techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03 num_bytes: 614302 num_examples: 3735 - name: techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10 num_bytes: 608645 num_examples: 3735 - name: techiaith__kaldi_cy__main num_bytes: 589751 num_examples: 3735 - name: mistralai__Voxtral_Mini_3B_2507__main num_bytes: 645993 num_examples: 3735 - name: techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10 num_bytes: 573910 num_examples: 3735 - name: openai__whisper_large_v2__main num_bytes: 640778 num_examples: 3735 download_size: 2643576 dataset_size: 4315439 - config_name: techiaith__banc_trawsgrifiadau_bangor features: - name: sentence dtype: string - name: id dtype: string - name: prediction dtype: string splits: - name: openai__whisper_large_v3__main num_bytes: 557956 num_examples: 3899 - name: techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03 num_bytes: 512782 num_examples: 3899 - name: techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10 num_bytes: 513324 num_examples: 3899 - name: techiaith__kaldi_cy__main num_bytes: 505425 num_examples: 3899 - name: mistralai__Voxtral_Mini_3B_2507__main num_bytes: 562286 num_examples: 3899 - name: techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10 num_bytes: 493366 num_examples: 3899 - name: openai__whisper_large_v2__main num_bytes: 548171 num_examples: 3899 download_size: 2565257 dataset_size: 3693310 - config_name: techiaith__commonvoice_18_0_cy features: - name: sentence dtype: string - name: language dtype: string - name: id dtype: string - name: prediction dtype: string splits: - name: openai__whisper_large_v3__main num_bytes: 721977 num_examples: 5386 - name: techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03 num_bytes: 703208 num_examples: 5386 - name: techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10 num_bytes: 704248 num_examples: 5386 - name: techiaith__kaldi_cy__main num_bytes: 694893 num_examples: 5386 - name: mistralai__Voxtral_Mini_3B_2507__main num_bytes: 716742 num_examples: 5386 - name: techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10 num_bytes: 694167 num_examples: 5386 - name: openai__whisper_large_v2__main num_bytes: 712423 num_examples: 5386 download_size: 3439680 dataset_size: 4947658 - config_name: techiaith__commonvoice_18_0_cy_en features: - name: sentence dtype: string - name: language dtype: string - name: id dtype: string - name: prediction dtype: string splits: - name: openai__whisper_large_v3__main num_bytes: 1623825 num_examples: 10773 - name: techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03 num_bytes: 1575659 num_examples: 10773 - name: techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10 num_bytes: 1655514 num_examples: 10773 - name: techiaith__kaldi_cy__main num_bytes: 1490373 num_examples: 10773 - name: mistralai__Voxtral_Mini_3B_2507__main num_bytes: 1563785 num_examples: 10773 - name: techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10 num_bytes: 1465939 num_examples: 10773 - name: openai__whisper_large_v2__main num_bytes: 1639159 num_examples: 10773 download_size: 7587225 dataset_size: 11014254 configs: - config_name: cymen_arfor__lleisiau_arfor data_files: - split: openai__whisper_large_v3__main path: cymen_arfor__lleisiau_arfor/openai__whisper_large_v3__main-* - split: techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03 path: cymen_arfor__lleisiau_arfor/techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03-* - split: techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10 path: cymen_arfor__lleisiau_arfor/techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10-* - split: techiaith__kaldi_cy__main path: cymen_arfor__lleisiau_arfor/techiaith__kaldi_cy__main-* - split: mistralai__Voxtral_Mini_3B_2507__main path: cymen_arfor__lleisiau_arfor/mistralai__Voxtral_Mini_3B_2507__main-* - split: techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10 path: cymen_arfor__lleisiau_arfor/techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10-* - split: openai__whisper_large_v2__main path: cymen_arfor__lleisiau_arfor/openai__whisper_large_v2__main-* - config_name: techiaith__banc_trawsgrifiadau_bangor data_files: - split: openai__whisper_large_v3__main path: techiaith__banc_trawsgrifiadau_bangor/openai__whisper_large_v3__main-* - split: techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03 path: techiaith__banc_trawsgrifiadau_bangor/techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03-* - split: techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10 path: techiaith__banc_trawsgrifiadau_bangor/techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10-* - split: techiaith__kaldi_cy__main path: techiaith__banc_trawsgrifiadau_bangor/techiaith__kaldi_cy__main-* - split: mistralai__Voxtral_Mini_3B_2507__main path: techiaith__banc_trawsgrifiadau_bangor/mistralai__Voxtral_Mini_3B_2507__main-* - split: techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10 path: techiaith__banc_trawsgrifiadau_bangor/techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10-* - split: openai__whisper_large_v2__main path: techiaith__banc_trawsgrifiadau_bangor/openai__whisper_large_v2__main-* - config_name: techiaith__commonvoice_18_0_cy data_files: - split: openai__whisper_large_v3__main path: techiaith__commonvoice_18_0_cy/openai__whisper_large_v3__main-* - split: techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03 path: techiaith__commonvoice_18_0_cy/techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03-* - split: techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10 path: techiaith__commonvoice_18_0_cy/techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10-* - split: techiaith__kaldi_cy__main path: techiaith__commonvoice_18_0_cy/techiaith__kaldi_cy__main-* - split: mistralai__Voxtral_Mini_3B_2507__main path: techiaith__commonvoice_18_0_cy/mistralai__Voxtral_Mini_3B_2507__main-* - split: techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10 path: techiaith__commonvoice_18_0_cy/techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10-* - split: openai__whisper_large_v2__main path: techiaith__commonvoice_18_0_cy/openai__whisper_large_v2__main-* - config_name: techiaith__commonvoice_18_0_cy_en data_files: - split: openai__whisper_large_v3__main path: techiaith__commonvoice_18_0_cy_en/openai__whisper_large_v3__main-* - split: techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03 path: techiaith__commonvoice_18_0_cy_en/techiaith__whisper_large_v3_ft_verbatim_cy_en__25.03-* - split: techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10 path: techiaith__commonvoice_18_0_cy_en/techiaith__whisper_large_v3_ft_verbatim_cy_en__24.10-* - split: techiaith__kaldi_cy__main path: techiaith__commonvoice_18_0_cy_en/techiaith__kaldi_cy__main-* - split: mistralai__Voxtral_Mini_3B_2507__main path: techiaith__commonvoice_18_0_cy_en/mistralai__Voxtral_Mini_3B_2507__main-* - split: techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10 path: techiaith__commonvoice_18_0_cy_en/techiaith__wav2vec2_btb_cv_ft_cv_cy__24.10-* - split: openai__whisper_large_v2__main path: techiaith__commonvoice_18_0_cy_en/openai__whisper_large_v2__main-* --- # Welsh ASR Model Evaluation Transcription Dataset This resource compiles the output transcriptions from multiple Welsh Automatic Speech Recognition (ASR) models across several test sets. The data is structured hierarchically: - Splits delineate the individual test sets. - Configs within each split detail the performance (transcriptions) of a specific ASR model and its version on that set. # Metrics Results | model | test | wer | cer | | --- | --- | --- | --- | | openai/whisper-large-v2 | cymen-arfor/lleisiau-arfor | 71.8988 | 49.4113 | | openai/whisper-large-v3 | cymen-arfor/lleisiau-arfor | 78.381 | 51.8036 | | techiaith/whisper-large-v3-ft-verbatim-cy-en/25.03 | cymen-arfor/lleisiau-arfor | 26.2495 | 9.9973 | | techiaith/whisper-large-v3-ft-verbatim-cy-en/24.10 | cymen-arfor/lleisiau-arfor | 49.9122 | 20.3199 | | techiaith/kaldi-cy | cymen-arfor/lleisiau-arfor | 54.5076 | 27.4573 | | mistralai/Voxtral-Mini-3B-2507 | cymen-arfor/lleisiau-arfor | 105.7659 | 76.3593 | | techiaith/wav2vec2-btb-cv-ft-cv-cy/24.10 | cymen-arfor/lleisiau-arfor | 68.31 | 33.8774 | | openai/whisper-large-v2 | techiaith/banc-trawsgrifiadau-bangor | 77.0912 | 50.902 | | openai/whisper-large-v3 | techiaith/banc-trawsgrifiadau-bangor | 86.1427 | 57.2622 | | techiaith/whisper-large-v3-ft-verbatim-cy-en/25.03 | techiaith/banc-trawsgrifiadau-bangor | 27.9029 | 9.7063 | | techiaith/whisper-large-v3-ft-verbatim-cy-en/24.10 | techiaith/banc-trawsgrifiadau-bangor | 27.6264 | 9.7113 | | techiaith/kaldi-cy | techiaith/banc-trawsgrifiadau-bangor | 46.1966 | 21.8032 | | mistralai/Voxtral-Mini-3B-2507 | techiaith/banc-trawsgrifiadau-bangor | 116.6845 | 86.3502 | | techiaith/wav2vec2-btb-cv-ft-cv-cy/24.10 | techiaith/banc-trawsgrifiadau-bangor | 53.9631 | 21.3484 | | openai/whisper-large-v2 | techiaith/commonvoice-18-0-cy | 37.9589 | 15.3495 | | openai/whisper-large-v3 | techiaith/commonvoice-18-0-cy | 44.5461 | 20.2724 | | techiaith/whisper-large-v3-ft-verbatim-cy-en/25.03 | techiaith/commonvoice-18-0-cy | 15.5251 | 4.1814 | | techiaith/whisper-large-v3-ft-verbatim-cy-en/24.10 | techiaith/commonvoice-18-0-cy | 4.1731 | 1.2283 | | techiaith/kaldi-cy | techiaith/commonvoice-18-0-cy | 50.9511 | 17.8832 | | mistralai/Voxtral-Mini-3B-2507 | techiaith/commonvoice-18-0-cy | 70.4097 | 37.963 | | techiaith/wav2vec2-btb-cv-ft-cv-cy/24.10 | techiaith/commonvoice-18-0-cy | 38.2517 | 9.7546 | | openai/whisper-large-v2 | techiaith/commonvoice-18-0-cy-en | 38.1935 | 21.5529 | | openai/whisper-large-v3 | techiaith/commonvoice-18-0-cy-en | 39.1813 | 22.6869 | | techiaith/whisper-large-v3-ft-verbatim-cy-en/25.03 | techiaith/commonvoice-18-0-cy-en | 28.1116 | 15.1831 | | techiaith/whisper-large-v3-ft-verbatim-cy-en/24.10 | techiaith/commonvoice-18-0-cy-en | 50.7933 | 36.4355 | | techiaith/kaldi-cy | techiaith/commonvoice-18-0-cy-en | 81.3322 | 46.1831 | | mistralai/Voxtral-Mini-3B-2507 | techiaith/commonvoice-18-0-cy-en | 42.2153 | 21.7994 | | techiaith/wav2vec2-btb-cv-ft-cv-cy/24.10 | techiaith/commonvoice-18-0-cy-en | 69.696 | 36.1913 |