paper-with-me

Papers

Avoid Overthinking in Self-Supervised Models for Speech Recognition

2022-11-01 · Dan Berrebbi, Brian Yan, Shinji Watanabe

Self-supervised learning (SSL) models reshaped our approach to speech, language and vision. However their huge size and the opaque relations between their layers and tasks result in slow inference and network overthinking, where predictions made from the last layer of large models is worse than those made from intermediate layers. Early exit (EE) strategies can solve both issues by dynamically reducing computations at inference time for certain samples. Although popular for classification tasks in vision and language, EE has seen less use for sequence-to-sequence speech recognition (ASR) tasks where outputs from early layers are often degenerate. This challenge is further compounded when speech SSL models are applied on out-of-distribution (OOD) data. This paper first shows that SSL models do overthinking in ASR. We then motivate further research in EE by computing an optimal bound for performance versus speed trade-offs. To approach this bound we propose two new strategies for ASR: (1) we adapt the recently proposed patience strategy to ASR; and (2) we design a new EE strategy specific to ASR that performs better than all strategies previously introduced.

📄 PDF Abstract BibTeX arXiv:2211.08989

Code (0)

등록된 구현이 없습니다.

Tasks

Self-Supervised LearningSequence-To-Sequence Speech Recognitionspeech-recognitionSpeech Recognition

Methods 이 논문이 사용한 방법론

SPEED The monocular depth estimation (MDE) is the task of estimating depth from a single frame. This information is an essential knowledge in many computer vision tasks such as scene…

Similar Papers 제목 키워드 기반

Revisiting Overthinking in Long Chain-of-Thought from the Perspective of Self-Doubt

2025-05-29 · Keqin Peng, Liang Ding, Yuanxin Ouyang, Meng Fang 외

Reasoning Large Language Models (RLLMs) have demonstrated impressive performance on complex tasks, largely due to the adoption of Long Chain-of-Thought (Long CoT) reasoning. However, they often exhibit overthinking -- pe…

Mathematical Reasoning

Dropout Regularization for Self-Supervised Learning of Transformer Encoder Speech Representation

2021-07-09 · Jian Luo, Jianzong Wang, Ning Cheng, Jing Xiao

Predicting the altered acoustic frames is an effective way of self-supervised learning for speech representation. However, it is challenging to prevent the pretrained model from overfitting. In this paper, we proposed to…

ClassificationSelf-Supervised LearningSpeaker Recognition

A Noise-Robust Self-supervised Pre-training Model Based Speech Representation Learning for Automatic Speech Recognition

2022-01-22 · Qiu-Shi Zhu, Jie Zhang, Zi-Qiang Zhang, Ming-Hui Wu 외

Wav2vec2.0 is a popular self-supervised pre-training framework for learning speech representations in the context of automatic speech recognition (ASR). It was shown that wav2vec2.0 has a good robustness against the doma…

Automatic Speech RecognitionAutomatic Speech Recognition (ASR)Representation Learningspeech-recognition+2

Mandarin-English Code-switching Speech Recognition with Self-supervised Speech Representation Models

2021-10-07 · Liang-Hsuan Tseng, Yu-Kuan Fu, Heng-Jui Chang, Hung-Yi Lee

Code-switching (CS) is common in daily conversations where more than one language is used within a sentence. The difficulties of CS speech recognition lie in alternating languages and the lack of transcribed data. Theref…

Language IdentificationSelf-Supervised LearningSentencespeech-recognition+1

Unified Speech-Text Pre-training for Speech Translation and Recognition

2022-04-11 · ACL 2022 5 · Yun Tang, Hongyu Gong, Ning Dong, Changhan Wang 외

We describe a method to jointly pre-train speech and text in an encoder-decoder modeling framework for speech translation and recognition. The proposed method incorporates four self-supervised and supervised subtasks for…

Decoderspeech-recognitionSpeech RecognitionTranslation