Automatic Speech Recognition

Automatic Speech Recognition (ASR) aims to accurately transcribe spoken language into text, driving research into robust and efficient models. Current efforts focus on improving accuracy and robustness through techniques like consistency regularization in Connectionist Temporal Classification (CTC), leveraging pre-trained multilingual models for low-resource languages, and integrating Large Language Models (LLMs) for enhanced contextual understanding and improved handling of diverse accents and speech disorders. These advancements have significant implications for accessibility, enabling applications in diverse fields such as healthcare, education, and human-computer interaction.

Papers

October 31, 2022

October 30, 2022

Improvements to Embedding-Matching Acoustic-to-Word ASR Using Multiple-Hypothesis Pronunciation-Based Embeddings
Hao Yen, Woojay Jeon
Automatic Speech Recognition Large Relevance Improvement Acoustic Word Embeddings Phonetic Embeddings Depth Hypothesis

October 27, 2022

October 26, 2022

Automatic Speech Recognition

Papers

Delay-penalized transducer for low-latency streaming ASR

Predicting Multi-Codebook Vector Quantization Indexes for Knowledge Distillation

Structured State Space Decoder for Speech Recognition and Synthesis

Blank Collapse: Compressing CTC emission for the faster decoding

Improvements to Embedding-Matching Acoustic-to-Word ASR Using Multiple-Hypothesis Pronunciation-Based Embeddings

Simulating realistic speech overlaps improves multi-talker ASR

Exploring Effective Distillation of Self-Supervised Speech Models for Automatic Speech Recognition

Make More of Your Data: Minimal Effort Data Augmentation for Automatic Speech Recognition and Translation

Robust Data2vec: Noise-robust Speech Representation Learning for ASR by Combining Regression and Improved Contrastive Learning

On Out-of-Distribution Detection for Audio with Deep Nearest Neighbors

Weight Averaging: A Simple Yet Effective Method to Overcome Catastrophic Forgetting in Automatic Speech Recognition

Contextual-Utterance Training for Automatic Speech Recognition

Streaming Voice Conversion Via Intermediate Bottleneck Features And Non-streaming Teacher Guidance

TRScore: A Novel GPT-based Readability Scorer for ASR Segmentation and Punctuation model evaluation and selection

Four-in-One: A Joint Approach to Inverse Text Normalization, Punctuation, Capitalization, and Disfluency for Automatic Speech Recognition

There is more than one kind of robustness: Fooling Whisper with adversarial examples

Efficient Utilization of Large Pre-Trained Models for Low Resource ASR

Monotonic segmental attention for automatic speech recognition

Reducing Language confusion for Code-switching Speech Recognition with Token-level Language Diarization

UFO2: A unified pre-training framework for online and offline speech recognition