arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2305.18596cs.SDcs.CLcs.LGeess.AS
收藏

为流式语音搜索构建准确的低延迟自动语音识别系统

Building Accurate Low Latency ASR for Streaming Voice Search

Flipkart(Flipkart(弗利普卡特))

Abhinav Goyal, Nikesh Garera

AI总结:

本文针对大规模 Hinglish 流式语音搜索,通过改进 LSTM 训练并联合 CTC 实现 ASR 与语音结束检测,在 Flipkart 数据上将词错误率降至 3.69%–4.78%,搜索延迟降低约 46.64%。

摘要:

自动语音识别(ASR)在基于语音的应用中起着至关重要的作用。对于像语音搜索这样需要实时反馈的应用,流式处理能力变得至关重要。虽然基于 LSTM/RNN 和 CTC 的 ASR 系统通常被用于低延迟流式应用,但由于缺乏未来音频帧,它们与最先进模型相比往往表现出较低的准确率。在这项工作中,我们专注于为大规模 Hinglish(印地语和英语的混合语言)语音搜索开发准确的基于 LSTM、注意力机制和 CTC 的流式 ASR 模型。我们研究了 vanilla LSTM 训练中的各种修改,这些修改在保持其流式处理能力的同时提高了系统的准确率。我们还解决了流式应用中语音结束(EOS)检测的关键需求。我们提出了一种用于端到端 CTC 模型的简单训练和推理策略,该策略能够实现 ASR 和 EOS 检测的联合执行。我们的模型在 Flipkart 语音搜索上的评估——该服务每天处理约 600 万次查询的庞大流量——表明相对于 vanilla LSTM-CTC 模型取得了显著的性能提升。我们的模型在没有 EOS 的情况下实现了 3.69% 的词错误率(WER),在有 EOS 的情况下实现了 4.78% 的词错误率,同时与独立的语音活动检测(VAD)模型相比,搜索延迟降低了约 1300 毫秒(相当于降低了 46.64%)。

英文摘要:

Automatic Speech Recognition (ASR) plays a crucial role in voice-based applications. For applications requiring real-time feedback like Voice Search, streaming capability becomes vital. While LSTM/RNN and CTC based ASR systems are commonly employed for low-latency streaming applications, they often exhibit lower accuracy compared to state-of-the-art models due to a lack of future audio frames. In this work, we focus on developing accurate LSTM, attention, and CTC based streaming ASR models for large-scale Hinglish (a blend of Hindi and English) Voice Search. We investigate various modifications in vanilla LSTM training which enhance the system's accuracy while preserving its streaming capabilities. We also address the critical requirement of end-of-speech (EOS) detection in streaming applications. We present a simple training and inference strategy for end-to-end CTC models that enables joint ASR and EOS detection. The evaluation of our model on Flipkart's Voice Search, which handles substantial traffic of approximately 6 million queries per day, demonstrates significant performance gains over the vanilla LSTM-CTC model. Our model achieves a word error rate (WER) of 3.69% without EOS and 4.78% with EOS while also reducing the search latency by approximately ~1300 ms (equivalent to 46.64% reduction) when compared to an independent voice activity detection (VAD) model.

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。

↑