الرئيسية المنصة الوثائق الأخبار الأوراق البحثية الدروس مجموعات البيانات الموسوعة SOTA نماذج LLM لوحة الأداء GPU الفعاليات

العربية

Speech Recognition On Wsj Eval92

المقاييس

Word Error Rate (WER)

النتائج

نتائج أداء النماذج المختلفة على هذا المعيار القياسي

		Paper Title	Repository
Jasper 10x3	6.9	Jasper: An End-to-End Convolutional Neural Acoustic Model
CNN over RAW speech (wav)	5.6	-	-
CTC-CRF 4gram-LM	3.79	CRF-based Single-stage Acoustic Modeling with CTC Topology	-
test-set on open vocabulary (i.e. harder), model = HMM-DNN + pNorm*	3.6	-	-
Deep Speech 2	3.60	Deep Speech 2: End-to-End Speech Recognition in English and Mandarin
TC-DNN-BLSTM-DNN	3.5	Deep Recurrent Neural Networks for Acoustic Modelling	-
Convolutional Speech Recognition	3.5	Fully Convolutional Speech Recognition	-
Espresso	3.4	Espresso: A Fast End-to-end Neural Speech Recognition Toolkit
CTC-CRF VGG-BLSTM	3.2	CAT: A CTC-CRF based ASR Toolkit Bridging the Hybrid and the End-to-end Approaches towards Data Efficiency and Low Latency
Transformer with Relaxed Attention	3.19	Relaxed Attention: A Simple Method to Boost Performance of End-to-End Automatic Speech Recognition
End-to-end LF-MMI	3.0	End-to-end speech recognition using lattice-free MMI	-
CTC-CRF ST-NAS	2.77	Efficient Neural Architecture Search for End-to-end Speech Recognition via Straight-Through Gradients
tdnn + chain	2.32	Purely sequence-trained neural networks for ASR based on lattice-free MMI	-
RobustGER	2.2	It's Never Too Late: Fusing Acoustic Information into Large Language Models for Automatic Speech Recognition
Task activating prompting generative correction	2.11	Generative Speech Recognition Error Correction with Large Language Models and Task-Activating Prompting	-
ConformerXXL-P	1.3	BigSSL: Exploring the Frontier of Large-Scale Semi-Supervised Learning for Automatic Speech Recognition	-
Speechstew 100M	1.3	SpeechStew: Simply Mix All Available Speech Recognition Data to Train One Large Neural Network	-

0 of 17 row(s) selected.