Towards Fine-Grained and Multi-Granular Contrastive Language-Speech Pre-training
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915942705922048 |
|---|---|
| author | Yang, Yifan Han, Bing Wang, Hui Wang, Wei Ma, Ziyang Zhou, Long Jin, Zengrui Yang, Guanrou Wang, Tianrui Tan, Xu Chen, Xie |
| author_facet | Yang, Yifan Han, Bing Wang, Hui Wang, Wei Ma, Ziyang Zhou, Long Jin, Zengrui Yang, Guanrou Wang, Tianrui Tan, Xu Chen, Xie |
| contents | Modeling fine-grained speaking styles remains challenging for language-speech representation pre-training, as existing speech-text models are typically trained with coarse captions or task-specific supervision, and scalable fine-grained style annotations are unavailable. We present FCaps, a large-scale dataset with fine-grained free-text style descriptions, encompassing 47k hours of speech and 19M fine-grained captions annotated via a novel end-to-end pipeline that directly grounds detailed captions in audio, thereby avoiding the error propagation caused by LLM-based rewriting in existing cascaded pipelines. Evaluations using LLM-as-a-judge demonstrate that our annotations surpass existing cascaded annotations in terms of correctness, coverage, and naturalness. Building on FCaps, we propose CLSP, a contrastive language-speech pre-trained model that integrates global and fine-grained supervision, enabling unified representations across multiple granularities. Extensive experiments demonstrate that CLSP learns fine-grained and multi-granular speech-text representations that perform reliably across global and fine-grained speech-text retrieval, zero-shot paralinguistic classification, and speech style similarity scoring, with strong alignment to human judgments. Code and dataset are publicly available at https://github.com/yfyeung/CLSP. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_03065 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Towards Fine-Grained and Multi-Granular Contrastive Language-Speech Pre-training Yang, Yifan Han, Bing Wang, Hui Wang, Wei Ma, Ziyang Zhou, Long Jin, Zengrui Yang, Guanrou Wang, Tianrui Tan, Xu Chen, Xie Audio and Speech Processing Modeling fine-grained speaking styles remains challenging for language-speech representation pre-training, as existing speech-text models are typically trained with coarse captions or task-specific supervision, and scalable fine-grained style annotations are unavailable. We present FCaps, a large-scale dataset with fine-grained free-text style descriptions, encompassing 47k hours of speech and 19M fine-grained captions annotated via a novel end-to-end pipeline that directly grounds detailed captions in audio, thereby avoiding the error propagation caused by LLM-based rewriting in existing cascaded pipelines. Evaluations using LLM-as-a-judge demonstrate that our annotations surpass existing cascaded annotations in terms of correctness, coverage, and naturalness. Building on FCaps, we propose CLSP, a contrastive language-speech pre-trained model that integrates global and fine-grained supervision, enabling unified representations across multiple granularities. Extensive experiments demonstrate that CLSP learns fine-grained and multi-granular speech-text representations that perform reliably across global and fine-grained speech-text retrieval, zero-shot paralinguistic classification, and speech style similarity scoring, with strong alignment to human judgments. Code and dataset are publicly available at https://github.com/yfyeung/CLSP. |
| title | Towards Fine-Grained and Multi-Granular Contrastive Language-Speech Pre-training |
| topic | Audio and Speech Processing |
| url | https://arxiv.org/abs/2601.03065 |