SOTAVerified

Speech Recognition

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.

( Image credit: SpecAugment )

Papers

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 1751–1800 of 6433 papers

TitleStatusHype
Exploring Speech Recognition, Translation, and Understanding with Discrete Speech Units: A Comparative Study—0
Generative Speech Recognition Error Correction with Large Language Models and Task-Activating Prompting—0
Low-rank Adaptation of Large Language Model Rescoring for Parameter-Efficient Speech Recognition—0
Segmentation-Free Streaming Machine TranslationCode0
Segment-Level Vectorized Beam Search Based on Partially Autoregressive Inference—0
Learning from Flawed Data: Weakly Supervised Automatic Speech Recognition—0
Unsupervised Pre-Training for Vietnamese Automatic Speech Recognition in the HYKIST Project—0
AutoPrep: An Automatic Preprocessing Framework for In-the-Wild Speech Data—0
On the Relation between Internal Language Model and Sequence Discriminative Training for Neural Transducers—0
Connecting Speech Encoder and Large Language Model for ASR—0
On the Impact of Quantization and Pruning of Self-Supervised Speech Models for Downstream Speech Recognition Tasks "In-the-Wild''—0
Unsupervised Accent Adaptation Through Masked Language Model Correction Of Discrete Self-Supervised Speech Units—0
Cross-modal Alignment with Optimal Transport for CTC-based ASR—0
Speech enhancement with frequency domain auto-regressive modeling—0
Human Transcription Quality ImprovementCode0
My Science Tutor (MyST) -- A Large Corpus of Children's Conversational Speech—0
NTT speaker diarization system for CHiME-7: multi-domain, multi-microphone End-to-end and vector clustering diarization—0
Importance of Smoothness Induced by Optimizers in FL4ASR: Towards Understanding Federated Learning for End-to-End ASR—0
Dynamic ASR Pathways: An Adaptive Masking Approach Towards Efficient Pruning of A Multilingual ASR Model—0
Affect Recognition in Conversations Using Large Language Models—0
Massive End-to-end Models for Short Search Queries—0
Big model only for hard audios: Sample dependent Whisper model selection for efficient inferencesCode0
Sparsely Shared LoRA on Whisper for Child Speech Recognition—0
A Multiscale Autoencoder (MSAE) Framework for End-to-End Neural Network Speech Enhancement—0
CoMFLP: Correlation Measure based Fast Search on ASR Layer PruningCode0
Variational Connectionist Temporal Classification for Order-Preserving Sequence Modeling—0
AudioFool: Fast, Universal and synchronization-free Cross-Domain Attack on Speech Recognition—0
Leveraging Data Collection and Unsupervised Learning for Code-switched Tunisian Arabic Automatic Speech Recognition—0
End-to-End Speech Recognition Contextualization with Large Language Models—0
Incorporating Ultrasound Tongue Images for Audio-Visual Speech Enhancement—0
Semi-Autoregressive Streaming ASR With Label Context—0
Discrete Audio Representation as an Alternative to Mel-Spectrograms for Speaker and Speech Recognition—0
Harnessing the Zero-Shot Power of Instruction-Tuned Large Language Model in End-to-End Speech Recognition—0
Exploring Speech Enhancement for Low-resource Speech Synthesis—0
Enhancing Multilingual Speech Recognition through Language Prompt Tuning and Frame-Level Language Adapter—0
Corpus Synthesis for Zero-shot ASR domain Adaptation using Large Language Models—0
Investigating End-to-End ASR Architectures for Long Form Audio Transcription—0
Training dynamic models using early exits for automatic speech recognition on resource-constrained devicesCode0
Distilling HuBERT with LSTMs via Decoupled Knowledge Distillation—0
Instruction-Following Speech Recognition—0
HTEC: Human Transcription Error Correction—0
A Multitask Training Approach to Enhance Whisper with Contextual Biasing and Open-Vocabulary Keyword Spotting—0
Enhancing Quantised End-to-End ASR Models via PersonalisationCode0
Continuous Modeling of the Denoising Process for Speech Enhancement Based on Deep Learning—0
Decoder-only Architecture for Speech Recognition with CTC Prompts and Text Data Augmentation—0
Improving Speech Recognition for African American English With Audio Classification—0
Boosting End-to-End Multilingual Phoneme Recognition through Exploiting Universal Speech Attributes Constraints—0
Combining TF-GridNet and Mixture Encoder for Continuous Speech Separation for Meeting Transcription—0
Augmenting conformers with structured state-space sequence models for online speech recognition—0
Transformer Based Punctuation Restoration for TurkishCode0
Show:102550
← PrevPage 36 of 129Next →
All datasetsLibriSpeech test-cleanLibriSpeech test-otherSwitchboard + Hub500TIMITAISHELL-1WSJ eval92Common Voice Germanswb_hub_500 WER fullSWBCHTUDACommon Voice FrenchCommon Voice SpanishMediaSpeech

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1AmNetWord Error Rate (WER)8.6—Unverified
2HMM-(SAT)GMMWord Error Rate (WER)8—Unverified
3Local Prior Matching (Large Model)Word Error Rate (WER)7.19—Unverified
4SnipsWord Error Rate (WER)6.4—Unverified
5Li-GRUWord Error Rate (WER)6.2—Unverified
6HMM-DNN + pNorm*Word Error Rate (WER)5.5—Unverified
7CTC + policy learningWord Error Rate (WER)5.42—Unverified
8Deep Speech 2Word Error Rate (WER)5.33—Unverified
9Gated ConvNetsWord Error Rate (WER)4.8—Unverified
10HMM-TDNN + iVectorsWord Error Rate (WER)4.8—Unverified
#ModelMetricClaimedVerifiedStatus
1Local Prior Matching (Large Model)Word Error Rate (WER)20.84—Unverified
2SnipsWord Error Rate (WER)16.5—Unverified
3Local Prior Matching (Large Model, ConvLM LM)Word Error Rate (WER)15.28—Unverified
4Deep Speech 2Word Error Rate (WER)13.25—Unverified
5TDNN + pNorm + speed up/down speechWord Error Rate (WER)12.5—Unverified
6CTC-CRF 4gram-LMWord Error Rate (WER)10.65—Unverified
7Convolutional Speech RecognitionWord Error Rate (WER)10.47—Unverified
8MT4SSLWord Error Rate (WER)9.6—Unverified
9Jasper DR 10x5Word Error Rate (WER)8.79—Unverified
10EspressoWord Error Rate (WER)8.7—Unverified
#ModelMetricClaimedVerifiedStatus
1Deep SpeechPercentage error20—Unverified
2DNN-HMMPercentage error18.5—Unverified
3CD-DNNPercentage error16.1—Unverified
4DNNPercentage error16—Unverified
5DNN + DropoutPercentage error15—Unverified
6DNN BMMIPercentage error12.9—Unverified
7HMM-TDNN + pNorm + speed up/down speechPercentage error12.9—Unverified
8DNN MPEPercentage error12.9—Unverified
9DNN MMIPercentage error12.9—Unverified
10CNN + Bi-RNN + CTC (speech to letters), 25.9% WER if trainedonlyon SWBPercentage error12.6—Unverified
#ModelMetricClaimedVerifiedStatus
1LSNNPercentage error33.2—Unverified
2LAS multitask with indicators samplingPercentage error20.4—Unverified
3Soft Monotonic Attention (ours, offline)Percentage error20.1—Unverified
4QCNN-10L-256FMPercentage error19.64—Unverified
5Bi-LSTM + skip connections w/ CTCPercentage error17.7—Unverified
6Bi-RNN + AttentionPercentage error17.6—Unverified
7RNN-CRF on 24(x3) MFSCPercentage error17.3—Unverified
8CNN in time and frequency + dropout, 17.6% w/o dropoutPercentage error16.7—Unverified
9Light Gated Recurrent UnitsPercentage error16.7—Unverified
10GRUPercentage error16.6—Unverified
#ModelMetricClaimedVerifiedStatus
1AttWord Error Rate (WER)18.7—Unverified
2CTC/AttWord Error Rate (WER)6.7—Unverified
3BRA-EWord Error Rate (WER)6.63—Unverified
4CTC-CRF 4gram-LMWord Error Rate (WER)6.34—Unverified
5BATWord Error Rate (WER)4.97—Unverified
6ParaformerWord Error Rate (WER)4.95—Unverified
7U2Word Error Rate (WER)4.72—Unverified
8UMAWord Error Rate (WER)4.7—Unverified
9Lightweight TransducerWord Error Rate (WER)4.31—Unverified
10CIF-HKD With LMWord Error Rate (WER)4.1—Unverified
#ModelMetricClaimedVerifiedStatus
1Jasper 10x3Word Error Rate (WER)6.9—Unverified
2CNN over RAW speech (wav)Word Error Rate (WER)5.6—Unverified
3CTC-CRF 4gram-LMWord Error Rate (WER)3.79—Unverified
4Deep Speech 2Word Error Rate (WER)3.6—Unverified
5test-set on open vocabulary (i.e. harder), model = HMM-DNN + pNorm*Word Error Rate (WER)3.6—Unverified
6TC-DNN-BLSTM-DNNWord Error Rate (WER)3.5—Unverified
7Convolutional Speech RecognitionWord Error Rate (WER)3.5—Unverified
8EspressoWord Error Rate (WER)3.4—Unverified
9CTC-CRF VGG-BLSTMWord Error Rate (WER)3.2—Unverified
10Transformer with Relaxed AttentionWord Error Rate (WER)3.19—Unverified