--- tags: - espnet - audio - automatic-speech-recognition language: en datasets: - slurp_entity license: cc-by-4.0 --- ## ESPnet2 ASR model ### `pyf98/slurp_entity_e_branchformer` This model was trained by Yifan Peng using slurp_entity recipe in [espnet](https://github.com/espnet/espnet/). References: - [E-Branchformer: Branchformer with Enhanced merging for speech recognition (SLT 2022)](https://arxiv.org/abs/2210.00077) - [Branchformer: Parallel MLP-Attention Architectures to Capture Local and Global Context for Speech Recognition and Understanding (ICML 2022)](https://proceedings.mlr.press/v162/peng22a.html) ### Demo: How to use in ESPnet2 Follow the [ESPnet installation instructions](https://espnet.github.io/espnet/installation.html) if you haven't done that already. ```bash cd espnet git checkout 4bbd29a40cc7e2259996d30c0c76d3d789c1153d pip install -e . cd egs2/slurp_entity/asr1 ./run.sh --skip_data_prep false --skip_train true --download_model pyf98/slurp_entity_e_branchformer ``` # RESULTS ## Environments - date: `Mon Feb 27 19:14:30 CST 2023` - python version: `3.9.15 (main, Nov 24 2022, 14:31:59) [GCC 11.2.0]` - espnet version: `espnet 202301` - pytorch version: `pytorch 1.13.1` - Git hash: `4bbd29a40cc7e2259996d30c0c76d3d789c1153d` - Commit date: `Sat Feb 25 21:54:03 2023 -0600` ## exp/asr_train_asr_e_branchformer_e12_mlp3072_linear1024_layerdrop_raw_en_word ### WER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |decode_asr_asr_model_valid.acc.ave_10best/devel|8690|178058|84.6|7.6|7.8|3.2|18.6|51.2| |decode_asr_asr_model_valid.acc.ave_10best/test|13078|262176|83.7|7.7|8.6|3.0|19.3|49.7| ### CER |dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err| |---|---|---|---|---|---|---|---|---| |decode_asr_asr_model_valid.acc.ave_10best/devel|8690|847400|90.8|3.0|6.2|3.5|12.7|51.2| |decode_asr_asr_model_valid.acc.ave_10best/test|13078|1245475|89.7|3.1|7.2|3.4|13.6|49.7| ### Intent Classification - Valid Intent Classification Result: 0.8781357882623706 - Test Intent Classification Result: 0.8743691695977979 ### Entity |Slu f1|Precision|Recall|F-Measure| |:---:|:---:|:---:|:---:| | test | 0.7940 | 0.7582 | 0.7757 | ## ASR config
expand ``` config: conf/tuning/train_asr_e_branchformer_e12_mlp3072_linear1024_layerdrop.yaml print_config: false log_level: INFO dry_run: false iterator_type: sequence output_dir: exp/asr_train_asr_e_branchformer_e12_mlp3072_linear1024_layerdrop_raw_en_word ngpu: 1 seed: 0 num_workers: 1 num_att_plot: 3 dist_backend: nccl dist_init_method: env:// dist_world_size: null dist_rank: null local_rank: 0 dist_master_addr: null dist_master_port: null dist_launcher: null multiprocessing_distributed: false unused_parameters: false sharded_ddp: false cudnn_enabled: true cudnn_benchmark: false cudnn_deterministic: true collect_stats: false write_collected_feats: false max_epoch: 60 patience: null val_scheduler_criterion: - valid - loss early_stopping_criterion: - valid - loss - min best_model_criterion: - - valid - acc - max keep_nbest_models: 10 nbest_averaging_interval: 0 grad_clip: 5.0 grad_clip_type: 2.0 grad_noise: false accum_grad: 1 no_forward_run: false resume: true train_dtype: float32 use_amp: false log_interval: null use_matplotlib: true use_tensorboard: true create_graph_in_tensorboard: false use_wandb: false wandb_project: null wandb_id: null wandb_entity: null wandb_name: null wandb_model_log_interval: -1 detect_anomaly: false pretrain_path: null init_param: [] ignore_init_mismatch: false freeze_param: [] num_iters_per_epoch: null batch_size: 64 valid_batch_size: null batch_bins: 1000000 valid_batch_bins: null train_shape_file: - exp/asr_stats_raw_en_word/train/speech_shape - exp/asr_stats_raw_en_word/train/text_shape.word valid_shape_file: - exp/asr_stats_raw_en_word/valid/speech_shape - exp/asr_stats_raw_en_word/valid/text_shape.word batch_type: folded valid_batch_type: null fold_length: - 80000 - 150 sort_in_batch: descending sort_batch: descending multiple_iterator: false chunk_length: 500 chunk_shift_ratio: 0.5 num_cache_chunks: 1024 train_data_path_and_name_and_type: - - dump/raw/train/wav.scp - speech - kaldi_ark - - dump/raw/train/text - text - text valid_data_path_and_name_and_type: - - dump/raw/devel/wav.scp - speech - kaldi_ark - - dump/raw/devel/text - text - text allow_variable_data_keys: false max_cache_size: 0.0 max_cache_fd: 32 valid_max_cache_size: null exclude_weight_decay: false exclude_weight_decay_conf: {} optim: adam optim_conf: lr: 0.001 weight_decay: 1.0e-06 scheduler: warmuplr scheduler_conf: warmup_steps: 35000 token_list: - - - ▁SEP - ▁FILL - s - ▁the - a - ▁to - ▁i - ▁me - e - ▁s - ▁a - i - ▁you - ▁what - er - ing - u - ▁is - '''' - o - p - ▁in - ▁p - y - ▁my - ▁please - d - c - m - ▁b - l - ▁m - ▁c - st - date - n - ▁d - le - b - ▁for - re - t - ▁on - en - h - 'on' - ar - person - ▁re - ▁f - ▁g - ▁of - an - ▁ - g - ▁today - ▁t - or - ▁it - ▁this - ▁h - r - f - at - ch - ce - place_name - ▁email - ▁do - es - ri - ▁e - ▁w - ic - in - ▁that - event_name - ▁play - ▁and - al - ▁n - ▁can - email_query - ve - ▁new - day - it - ate - ▁from - ▁have - k - time - ▁am - media_type - email_sendemail - ent - ▁olly - qa_factoid - se - v - et - ck - ▁any - calendar_set - ly - th - ▁how - ▁meeting - ed - ▁tell - ▁st - x - ur - ro - ▁at - nd - ▁list - w - ▁u - ou - ▁not - ▁about - ▁an - ▁o - general_negate - ut - ▁time - ▁be - ▁ch - ▁are - social_post - business_name - la - ty - play_music - ot - general_quirky - ▁l - ▁sh - ▁tweet - om - ▁week - um - ▁one - ter - ▁he - ▁up - ▁com - general_praise - weather_query - ▁next - ▁th - ▁check - calendar_query - ▁last - ▁ro - ad - is - ▁with - ay - ▁send - pe - ▁pm - ▁tomorrow - ▁j - un - ▁train - general_explain - ▁v - one - ▁r - ra - news_query - ation - ▁emails - us - if - ct - ▁co - ▁add - ▁will - ▁se - nt - ▁was - ine - ▁de - ▁set - ▁ex - ▁would - ir - ow - ber - general_repeat - ight - ook - ▁again - ▁song - currency_name - ll - ▁ha - ▁go - relation - te - ion - and - ▁y - ▁ye - general_affirm - general_confirm - ery - ▁po - ff - ▁we - ▁turn - ▁did - ▁mar - ▁alarm - ▁like - datetime_query - ers - ▁all - ▁remind - ▁so - qa_definition - ▁calendar - end - ▁said - ci - ▁off - ▁john - ▁day - ss - pla - ume - ▁get - ail - pp - z - ry - am - ▁need - as - ▁thank - ▁wh - ▁want - ▁right - ▁jo - ▁facebook - ▁k - ge - ld - ▁fri - ▁two - general_dontcare - ▁news - ol - oo - ant - ▁five - ▁event - ake - definition_word - transport_type - ▁your - vi - orn - op - ▁weather - ome - ▁app - ▁lo - de - ▁music - weather_descriptor - ak - ke - ▁there - ▁si - ▁lights - ▁now - ▁mo - calendar_remove - our - ▁dollar - food_type - me - ▁more - ▁no - ▁birthday - orrect - ▁rep - ▁show - play_radio - ▁mon - ▁does - ood - ag - li - ▁sto - ▁contact - cket - email_querycontact - ▁ev - ▁could - ange - ▁just - out - ame - . - ▁ja - ▁confirm - qa_currency - ▁man - ▁late - ▁think - ▁some - timeofday - ▁bo - qa_stock - ong - ▁start - ▁work - ▁ten - int - ▁command - all - ▁make - ▁la - j - ▁answ - ▁hour - ▁cle - ah - ▁find - ▁service - ▁fa - qu - general_commandstop - ai - ▁when - ▁te - ▁by - social_query - ard - ▁tw - ul - id - ▁seven - ▁where - ▁much - art - ▁appointment - ver - artist_name - el - device_type - ▁know - ▁three - ▁events - ▁tr - ▁li - ork - red - ect - ▁let - ▁respon - ▁par - zz - ▁give - ▁twenty - ▁ti - ▁curre - play_podcasts - ▁radio - cooking_recipe - transport_query - ▁con - gh - ▁le - lists_query - ▁rem - recommendation_events - house_place - alarm_set - play_audiobook - ist - ase - music_genre - ive - ast - player_setting - ort - lly - news_topic - list_name - ▁playlist - ▁ne - business_type - personal_info - ind - ust - di - ress - recommendation_locations - lists_createoradd - iot_hue_lightoff - lists_remove - ord - ▁light - ere - alarm_query - audio_volume_mute - music_query - ▁audio - rain - ▁date - ▁order - audio_volume_up - ▁ar - ▁podcast - transport_ticket - mail - iot_hue_lightchange - iot_coffee - radio_name - ill - ▁ri - '@' - takeaway_query - song_name - takeaway_order - ▁ra - email_addcontact - play_game - book - transport_traffic - ▁house - music_likeness - her - transport_taxi - iot_hue_lightdim - ment - ght - fo - order_type - color_type - '1' - ven - ould - general_joke - ess - ain - qa_maths - ▁place - ▁twe - cast - iot_cleaning - ▁che - ▁cont - ith - audiobook_name - email_address - game_name - ▁cal - general_frequency - ▁tom - ▁food - act - iot_hue_lightup - '2' - alarm_remove - podcast_descriptor - ▁definition - audio_volume_down - ▁media - email_folder - dia - meal_type - ▁mus - recommendation_movies - ▁ad - ree - pt - now - playlist_name - ▁person - change_amount - ▁pla - escri - datetime_convert - podcast_name - ▁ab - time_zone - ▁def - ting - iot_wemo_on - music_settings - iot_wemo_off - orre - cy - ank - music_descriptor - lar - app_name - row - joke_type - xt - of - ition - ▁meet - ink - ▁confir - transport_agency - general_greet - ▁business - ▁art - ▁ag - urn - escript - rom - ▁rel - ▁au - ▁currency - audio_volume_other - iot_hue_lighton - ▁artist - '?' - ▁bus - cooking_type - movie_name - coffee_type - ingredient - ather - music_dislikeness - sp - q - ▁ser - esc - ▁bir - ▁cur - name - ▁tran - ▁hou - ek - uch - ▁conf - ▁face - '9' - ▁birth - I - sw - transport_descriptor - ▁comm - lease - transport_name - aid - movie_type - ▁device - alarm_type - audiobook_author - '5' - drink_type - ▁joh - ▁defin - word - ▁curren - order - iness - W - cooking_query - sport_type - ▁relation - oint - H - '8' - A - '0' - ▁dol - vice - ▁pers - '&' - T - ▁appoint - _ - '7' - '3' - '-' - game_type - ▁pod - N - M - E - list - music_album - dio - ▁transport - qa_query - C - O - U - query_detail - ']' - '[' - descriptor - ':' - spon - init: null input_size: null ctc_conf: dropout_rate: 0.0 ctc_type: builtin reduce: true ignore_nan_grad: null zero_infinity: true joint_net_conf: null use_preprocessor: true token_type: word bpemodel: null non_linguistic_symbols: null cleaner: null g2p: null speech_volume_normalize: null rir_scp: null rir_apply_prob: 1.0 noise_scp: null noise_apply_prob: 1.0 noise_db_range: '13_15' short_noise_thres: 0.5 aux_ctc_tasks: [] frontend: default frontend_conf: fs: 16k specaug: specaug specaug_conf: apply_time_warp: true time_warp_window: 5 time_warp_mode: bicubic apply_freq_mask: true freq_mask_width_range: - 0 - 30 num_freq_mask: 2 apply_time_mask: true time_mask_width_range: - 0 - 40 num_time_mask: 2 normalize: utterance_mvn normalize_conf: {} model: espnet model_conf: ctc_weight: 0.3 lsm_weight: 0.1 length_normalized_loss: false extract_feats_in_collect_stats: false preencoder: null preencoder_conf: {} encoder: e_branchformer encoder_conf: output_size: 512 attention_heads: 8 attention_layer_type: rel_selfattn pos_enc_layer_type: rel_pos rel_pos_type: latest cgmlp_linear_units: 3072 cgmlp_conv_kernel: 31 use_linear_after_conv: false gate_activation: identity num_blocks: 12 dropout_rate: 0.1 positional_dropout_rate: 0.1 attention_dropout_rate: 0.1 input_layer: conv2d layer_drop_rate: 0.1 linear_units: 1024 positionwise_layer_type: linear macaron_ffn: true use_ffn: true merge_conv_kernel: 31 postencoder: null postencoder_conf: {} decoder: transformer decoder_conf: attention_heads: 8 linear_units: 2048 num_blocks: 6 dropout_rate: 0.1 positional_dropout_rate: 0.1 self_attention_dropout_rate: 0.1 src_attention_dropout_rate: 0.1 layer_drop_rate: 0.2 preprocessor: default preprocessor_conf: {} required: - output_dir - token_list version: '202301' distributed: false ```
### Citing ESPnet ```BibTex @inproceedings{watanabe2018espnet, author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai}, title={{ESPnet}: End-to-End Speech Processing Toolkit}, year={2018}, booktitle={Proceedings of Interspeech}, pages={2207--2211}, doi={10.21437/Interspeech.2018-1456}, url={http://dx.doi.org/10.21437/Interspeech.2018-1456} } ``` or arXiv: ```bibtex @misc{watanabe2018espnet, title={ESPnet: End-to-End Speech Processing Toolkit}, author={Shinji Watanabe and Takaaki Hori and Shigeki Karita and Tomoki Hayashi and Jiro Nishitoba and Yuya Unno and Nelson Yalta and Jahn Heymann and Matthew Wiesner and Nanxin Chen and Adithya Renduchintala and Tsubasa Ochiai}, year={2018}, eprint={1804.00015}, archivePrefix={arXiv}, primaryClass={cs.CL} } ```