arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2406.00522eess.AScs.SD
收藏

Wav2Prompt:用于LLM在零样本和少样本学习中的端到端语音提示生成与微调

Wav2Prompt: End-to-End Speech Prompt Generation and Tuning For LLM in Zero and Few-shot Learning

University of Cambridge(剑桥大学)

Keqi Deng, Guangzhi Sun, Philip C. Woodland

AI总结:

本文提出Wav2Prompt框架,通过将语音转化为连续表示作为LLM提示,实现语音与文本LLM的端到端集成,在零样本任务中表现与ASR-LLM级联相当,在少样本微调场景下显著优于级联方法。

摘要:

本文提出了Wav2Prompt,它允许语音输入与基于文本的大型语言模型(LLM)直接集成。Wav2Prompt使用简单的训练过程,仅使用与训练自动语音识别(ASR)模型相同的数据。训练后,Wav2Prompt从语音中学习连续表示,并将其用作LLM提示。为避免以往工作中的任务过拟合问题并保留LLM的涌现能力,Wav2Prompt将LLM token embedding作为训练目标,并利用连续的integrate-and-fire机制进行显式的语音-文本对齐。因此,Wav2Prompt-LLM组合可应用于零样本语音语言任务,如语音翻译(ST)、语音理解(SLU)、语音问答(SQA)和基于语音查询的问答(SQQA)。实验表明,在这些零样本任务中,Wav2Prompt的表现与ASR-LLM级联相似,且优于近期的前期工作。如果在少样本场景中有少量特定任务的配对数据,Wav2Prompt-LLM组合可进行端到端(E2E)微调。对于上述任务,Wav2Prompt-LLM组合的结果显著优于ASR-LLM级联。例如,在使用BLOOMZ-7B1 LLM进行英法ST时,Wav2Prompt-LLM组合比ASR-LLM级联提高了8.5个BLEU点。

英文摘要:

Wav2Prompt is proposed which allows straightforward integration between spoken input and a text-based large language model (LLM). Wav2Prompt uses a simple training process with only the same data used to train an automatic speech recognition (ASR) model. After training, Wav2Prompt learns continuous representations from speech and uses them as LLM prompts. To avoid task over-fitting issues found in prior work and preserve the emergent abilities of LLMs, Wav2Prompt takes LLM token embeddings as the training targets and utilises a continuous integrate-and-fire mechanism for explicit speech-text alignment. Therefore, a Wav2Prompt-LLM combination can be applied to zero-shot spoken language tasks such as speech translation (ST), speech understanding (SLU), speech question answering (SQA) and spoken-query-based QA (SQQA). It is shown that for these zero-shot tasks, Wav2Prompt performs similarly to an ASR-LLM cascade and better than recent prior work. If relatively small amounts of task-specific paired data are available in few-shot scenarios, the Wav2Prompt-LLM combination can be end-to-end (E2E) fine-tuned. The Wav2Prompt-LLM combination then yields greatly improved results relative to an ASR-LLM cascade for the above tasks. For instance, for English-French ST with the BLOOMZ-7B1 LLM, a Wav2Prompt-LLM combination gave a 8.5 BLEU point increase over an ASR-LLM cascade.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑