PulseAugur
实时 03:28:39
English(EN) Leveraging Audio-LLMs to Filter Speech-to-Speech Training Data

Audio-LLMs 增强语音翻译数据过滤

研究人员开发了一种新颖的方法,利用音频大型语言模型 (Audio-LLMs) 来过滤嘈杂的语音到语音翻译 (S2ST) 训练数据。该方法采用两阶段的 Rank-to-Distill 策略,其中初始排序器生成用于保留或丢弃语音对的伪标签,然后训练一个 Audio-LLM 直接从音频中做出这些决策。该模型能有效捕捉声学保真度和跨语言语义一致性,从而在 S2ST 性能上取得显著改进,在基准数据集上 ASR-BLEU 的提升高达 +1.4。 AI

影响 提高了语音翻译模型的训练数据质量,有望带来更准确、更鲁棒的语音到语音翻译系统。

排序理由 该集群描述了一篇在 arXiv 上发表的研究论文,其中详细介绍了一种用于过滤语音到语音翻译训练数据的新方法。

在 arXiv cs.CL 阅读 →

AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →

Audio-LLMs 增强语音翻译数据过滤

本文如何被排名

Signal score
0 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Research
该集群描述了一篇在 arXiv 上发表的研究论文,其中详细介绍了一种用于过滤语音到语音翻译训练数据的新方法。
Source corroboration
2 independent sources
Multiple independent publishers reporting the same story raises confidence that it's real and newsworthy.
Topics
paper, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
80 days old
Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准

报道来源 [2]

  1. arXiv cs.CL TIER_1 English(EN) · Qixu Chen, Satoshi Nakamura ·

    利用 Audio-LLM 过滤语音到语音训练数据

    arXiv:2606.13507v1 Announce Type: new Abstract: Large-scale mined corpora provide abundant training data for end-to-end speech-to-speech translation (S2ST) but may contain noise, misalignment, and semantic errors. Filtering noisy data is crucial to maintain robust speech translat…

  2. arXiv cs.CL TIER_1 English(EN) · Satoshi Nakamura ·

    利用 Audio-LLM 过滤语音到语音训练数据

    Large-scale mined corpora provide abundant training data for end-to-end speech-to-speech translation (S2ST) but may contain noise, misalignment, and semantic errors. Filtering noisy data is crucial to maintain robust speech translation performance. We study how to train an audio-…