wav2vec2-large-xls-r-300m-spanish-small Open Source Model - Accurately Achieve Spanish Speech Recognition

Wav2vec2 Large Xls R 300m Spanish Small

Developed by tomascufaro

Spanish speech recognition model based on wav2vec2-large-xls-r-300m architecture, fine-tuned on the Common Voice dataset with a word error rate (WER) of 0.1791

Speech Recognition

Transformers

#Spanish speech recognition #Low word error rate #wav2vec2 fine-tuning

Downloads 13

Release Time : 3/2/2022

Model Overview

This is an automatic speech recognition (ASR) model optimized for Spanish, suitable for converting Spanish speech to text

Model Features

Low word error rate

Achieves a word error rate (WER) of 0.1791 on the evaluation set, demonstrating excellent performance

Large-scale pre-training

Based on the 300M-parameter wav2vec2 architecture with powerful speech feature extraction capabilities

Spanish optimization

Specifically fine-tuned for Spanish speech characteristics, adapting to Spanish pronunciation features

Model Capabilities

Spanish speech recognition

Speech-to-text

Continuous speech processing

Use Cases

Speech transcription

Meeting minutes

Automatically convert Spanish meeting recordings into text transcripts

Approximately 82% accuracy

Voice assistant

Speech recognition module for Spanish voice assistants

Education

Language learning applications

Help learners check the accuracy of Spanish pronunciation

🚀 wav2vec2-large-xls-r-300m-spanish-small

This model is a fine - tuned version of jhonparra18/wav2vec2-large-xls-r-300m-spanish-custom on the common_voice dataset. It offers high - quality speech recognition capabilities, achieving excellent results on the evaluation set.

🚀 Quick Start

This model is ready to use after fine - tuning. You can directly apply it to speech recognition tasks on the common_voice dataset.

📚 Documentation

Model Evaluation Results

It achieves the following results on the evaluation set:

Loss: 0.3763
Wer: 0.1791

Training and Evaluation Data

The model is fine - tuned on the common_voice dataset.

Training Procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0003
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 500
num_epochs: 30
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2277	0.26	400	0.2601	0.2291
0.2932	0.53	800	0.2950	0.2670
0.3019	0.79	1200	0.3247	0.2766
0.2987	1.05	1600	0.3031	0.2606
0.261	1.32	2000	0.2994	0.2620
0.2651	1.58	2400	0.3134	0.2700
0.264	1.85	2800	0.3016	0.2641
0.2475	2.11	3200	0.3135	0.2661
0.2269	2.37	3600	0.3029	0.2562
0.2389	2.64	4000	0.3035	0.2549
0.2319	2.9	4400	0.3022	0.2551
0.2123	3.16	4800	0.3256	0.2638
0.2094	3.43	5200	0.3227	0.2712
0.2121	3.69	5600	0.3085	0.2596
0.207	3.96	6000	0.3041	0.2597
0.1809	4.22	6400	0.3122	0.2524
0.1846	4.48	6800	0.3254	0.2579
0.1885	4.75	7200	0.2958	0.2437
0.1923	5.01	7600	0.3136	0.2502
0.1626	5.27	8000	0.3059	0.2488
0.1704	5.54	8400	0.3082	0.2515
0.1674	5.8	8800	0.3196	0.2509
0.1691	6.06	9200	0.3193	0.25
0.1499	6.33	9600	0.3529	0.2635
0.1568	6.59	10000	0.3241	0.2481
0.1538	6.86	10400	0.3354	0.2476
0.1503	7.12	10800	0.3180	0.2402
0.136	7.38	11200	0.3230	0.2397
0.1413	7.65	11600	0.3178	0.2451
0.147	7.91	12000	0.3170	0.2389
0.1341	8.17	12400	0.3380	0.2501
0.1329	8.44	12800	0.3265	0.2414
0.1314	8.7	13200	0.3281	0.2482
0.1312	8.97	13600	0.3259	0.2539
0.12	9.23	14000	0.3291	0.2424
0.1193	9.49	14400	0.3302	0.2412
0.1189	9.76	14800	0.3376	0.2407
0.1217	10.02	15200	0.3334	0.2400
0.1118	10.28	15600	0.3359	0.2368
0.1139	10.55	16000	0.3239	0.2335
0.1106	10.81	16400	0.3374	0.2352
0.1081	11.07	16800	0.3585	0.2434
0.1063	11.34	17200	0.3639	0.2472
0.1041	11.6	17600	0.3399	0.2423
0.1062	11.87	18000	0.3410	0.2388
0.1012	12.13	18400	0.3597	0.2413
0.0953	12.39	18800	0.3440	0.2296
0.097	12.66	19200	0.3440	0.2269
0.0968	12.92	19600	0.3498	0.2333
0.0902	13.18	20000	0.3471	0.2290
0.0868	13.45	20400	0.3462	0.2266
0.0892	13.71	20800	0.3373	0.2227
0.0902	13.97	21200	0.3377	0.2240
0.0846	14.24	21600	0.3484	0.2237
0.0839	14.5	22000	0.3706	0.2260
0.0834	14.77	22400	0.3430	0.2268
0.0841	15.03	22800	0.3489	0.2259
0.076	15.29	23200	0.3626	0.2281
0.0771	15.56	23600	0.3624	0.2268
0.0773	15.82	24000	0.3440	0.2252
0.0759	16.08	24400	0.3532	0.2170
0.0745	16.35	24800	0.3686	0.2188
0.0713	16.61	25200	0.3691	0.2195
0.0718	16.88	25600	0.3470	0.2108
0.0685	17.14	26000	0.3756	0.2179
0.0689	17.4	26400	0.3542	0.2149
0.0671	17.67	26800	0.3461	0.2165
0.0737	17.93	27200	0.3473	0.2238
0.0669	18.19	27600	0.3441	0.2138
0.0629	18.46	28000	0.3721	0.2155
0.0632	18.72	28400	0.3667	0.2126
0.0647	18.98	28800	0.3579	0.2097
0.0603	19.25	29200	0.3670	0.2130
0.0604	19.51	29600	0.3750	0.2142
0.0619	19.78	30000	0.3804	0.2160
0.0603	20.04	30400	0.3764	0.2124
0.0577	20.3	30800	0.3858	0.2097
0.0583	20.57	31200	0.3520	0.2089
0.0561	20.83	31600	0.3615	0.2079
0.0545	21.09	32000	0.3824	0.2032
0.0525	21.36	32400	0.3858	0.2091
0.0524	21.62	32800	0.3956	0.2099
0.0527	21.89	33200	0.3667	0.2025
0.0514	22.15	33600	0.3708	0.2032
0.0506	22.41	34000	0.3815	0.2053
0.0478	22.68	34400	0.3671	0.2007
0.049	22.94	34800	0.3758	0.2003
0.0477	23.2	35200	0.3786	0.2014
0.045	23.47	35600	0.3732	0.1998
0.0426	23.73	36000	0.3737	0.2010
0.0444	23.99	36400	0.3600	0.1990
0.0433	24.26	36800	0.3689	0.1976
0.0442	24.52	37200	0.3787	0.1968
0.0419	24.79	37600	0.3652	0.1961
0.042	25.05	38000	0.3820	0.1964
0.0419	25.31	38400	0.3786	0.1919
0.0376	25.58	38800	0.3842	0.1934
0.0385	25.84	39200	0.3767	0.1900
0.0396	26.1	39600	0.3688	0.1888
0.0371	26.37	40000	0.3815	0.1894
0.0363	26.63	40400	0.3748	0.1878
0.0377	26.9	40800	0.3713	0.1852
0.0352	27.16	41200	0.3734	0.1851
0.0355	27.42	41600	0.3776	0.1874
0.0333	27.69	42000	0.3867	0.1841
0.0348	27.95	42400	0.3823	0.1839
0.0329	28.21	42800	0.3795	0.1822
0.0325	28.48	43200	0.3711	0.1813
0.0328	28.74	43600	0.3721	0.1781
0.0312	29.0	44000	0.3803	0.1816
0.0318	29.27	44400	0.3758	0.1794
0.0302	29.53	44800	0.3792	0.1784
0.0339	29.8	45200	0.3763	0.1791

Framework versions

Transformers 4.16.0.dev0
Pytorch 1.10.1+cu102
Datasets 1.17.1.dev0
Tokenizers 0.11.0

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご