Data-Centric Lessons To Improve Speech-Language Pretraining

• URL: http://arxiv.org/abs/2510.20860v1
• Date: Wed, 22 Oct 2025 17:34:59 GMT
• Title: Data-Centric Lessons To Improve Speech-Language Pretraining
• Authors: Vishaal Udandarao, Zhiyun Lu, Xuankai Chang, Yongqiang Wang, Violet Z. Yao, Albin Madapally Jose, Fartash Faghri, Josh Gardner, Chung-Cheng Chiu,
• Abstract summary: Spoken Question-Answering (SQA) is a core capability for useful and interactive artificial intelligence systems.<n>We focus on three research questions fundamental to speech-language pretraining data.
• Score: 28.052057327597936
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Spoken Question-Answering (SQA) is a core capability for useful and interactive artificial intelligence systems. Recently, several speech-language models (SpeechLMs) have been released with a specific focus on improving their SQA performance. However, a lack of controlled ablations of pretraining data processing and curation makes it challenging to understand what factors account for performance, despite substantial gains from similar studies in other data modalities. In this work, we address this gap by conducting a data-centric exploration for pretraining SpeechLMs. We focus on three research questions fundamental to speech-language pretraining data: (1) how to process raw web-crawled audio content for speech-text pretraining, (2) how to construct synthetic pretraining datasets to augment web-crawled data and (3) how to interleave (text, audio) segments into training sequences. We apply the insights from our controlled data-centric ablations to pretrain a 3.8B-parameter SpeechLM, called SpeLangy, that outperforms models that are up to 3x larger by 10.2% absolute performance. We hope our findings highlight the impact of effective data curation for speech-language pretraining and guide future data-centric exploration in SpeechLMs.

Related papers

• Self-supervised learning of speech representations with Dutch archival data [8.504327926435158]
  We show how music, noise and speaker overlap affect SSL convergence and downstream fine-tuning performance.<n>We convert the noisy broadcast dataset into a qualitative dataset for pre-training, by using Whisper and WhisperX.<n>Finally, we achieve a state-of-the-art large wav2vec 2.0 model for the Dutch language, by a continuation of pre-training a wav2vec 2.0 XLS-R model checkpoint with our 55k hour archival dataset.
  arXiv  Detail & Related papers  (2025-07-06T22:11:22Z)
• Speech Unlearning [14.755831733659699]
  We introduce machine unlearning for speech tasks, a novel and underexplored research problem.<n>It aims to efficiently and effectively remove the influence of specific data from trained speech models without full retraining.<n>It has important applications in privacy preservation, removal of outdated or noisy data, and bias mitigation.
  arXiv  Detail & Related papers  (2025-06-01T06:04:16Z)
• Reasoning to Learn from Latent Thoughts [61.2395150828168]
  We show that explicitly modeling and inferring the emphlatent thoughts that underlie the text generation process can significantly improve pretraining data efficiency.<n>We show that a 1B LM can bootstrap its performance across at least three iterations and significantly outperform baselines trained on raw data.
  arXiv  Detail & Related papers  (2025-03-24T16:41:23Z)
• Scaling Speech-Text Pre-training with Synthetic Interleaved Data [31.77653849518526]
  Speech language models (SpeechLMs) accept speech input and produce speech output, allowing for more natural human-computer interaction.<n>Traditional approaches for developing SpeechLMs are constrained by the limited availability of unsupervised speech data and parallel speech-text data.<n>We propose a novel approach to scaling speech-text pre-training by leveraging large-scale synthetic interleaved data derived from text corpora.
  arXiv  Detail & Related papers  (2024-11-26T17:19:09Z)
• Reduce, Reuse, Recycle: Is Perturbed Data better than Other Language augmentation for Low Resource Self-Supervised Speech Models [48.44820587495038]
  Self-supervised representation learning (SSRL) has demonstrated superior performance than supervised models for tasks including phoneme recognition. Training SSRL models poses a challenge for low-resource languages where sufficient pre-training data may not be available. We propose to use audio augmentation techniques, namely: pitch variation, noise addition, accented target language and other language speech to pre-train SSRL models in a low resource condition and evaluate phoneme recognition.
  arXiv  Detail & Related papers  (2023-09-22T10:09:09Z)
• Textually Pretrained Speech Language Models [107.10344535390956]
  We propose TWIST, a method for training SpeechLMs using a warm-start from a pretrained textual language models. We show using both automatic and human evaluations that TWIST outperforms a cold-start SpeechLM across the board.
  arXiv  Detail & Related papers  (2023-05-22T13:12:16Z)
• SVTS: Scalable Video-to-Speech Synthesis [105.29009019733803]
  We introduce a scalable video-to-speech framework consisting of two components: a video-to-spectrogram predictor and a pre-trained neural vocoder. We are the first to show intelligible results on the challenging LRS3 dataset.
  arXiv  Detail & Related papers  (2022-05-04T13:34:07Z)
• Enhanced Direct Speech-to-Speech Translation Using Self-supervised Pre-training and Data Augmentation [76.13334392868208]
  Direct speech-to-speech translation (S2ST) models suffer from data scarcity issues. In this work, we explore self-supervised pre-training with unlabeled speech data and data augmentation to tackle this issue.
  arXiv  Detail & Related papers  (2022-04-06T17:59:22Z)
• Self-Supervised Learning for speech recognition with Intermediate layer supervision [52.93758711230248]
  We propose Intermediate Layer Supervision for Self-Supervised Learning (ILS-SSL) ILS-SSL forces the model to concentrate on content information as much as possible by adding an additional SSL loss on the intermediate layers. Experiments on LibriSpeech test-other set show that our method outperforms HuBERT significantly.
  arXiv  Detail & Related papers  (2021-12-16T10:45:05Z)
• A study on the efficacy of model pre-training in developing neural text-to-speech system [55.947807261757056]
  This study aims to understand better why and how model pre-training can positively contribute to TTS system performance. It is found that the TTS system could achieve comparable performance when the pre-training data is reduced to 1/8 of its original size.
  arXiv  Detail & Related papers  (2021-10-08T02:09:28Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.