arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2009.01008cs.CLcs.LGcs.SDeess.AS
收藏

带声学误差采样的跨话语语言模型

Cross-Utterance Language Models with Acoustic Error Sampling

Cambridge University(剑桥大学)

G. Sun, C. Zhang, P. C. Woodland

摘要:

如何有效地利用语言模型中丰富的上下文信息是自动语音识别的一个长期研究课题。本文提出了一种跨话语LM(CULM),它通过使用提取网络从过去和未来话语中获得上下文向量来增加标准长短期记忆(LSTM)LM的输入。提取网络使用另一个LSTM将周围的话语编码成向量,这些向量使用LSTM最终隐藏状态的投影或多头自关注层集成到上下文向量中。此外,提出了一种声学误差采样技术,以减少训练和测试时间之间的不匹配。这是通过将可能的ASR错误考虑到模型训练过程中来实现的,并且因此可以改善字出错率(WER)。在AMI和Switchboard数据集上进行的实验表明,CULM优于LSTM LM基线WER。特别是,CULM与自我关注的基于层的提取网络和声学误差采样实现了0.6%的绝对WER减少AMI,0.3%的WER减少交换机部分和0.9%的WER减少Eval2000测试集的Callhome部分在各自的基线。

英文摘要:

The effective exploitation of richer contextual information in language models (LMs) is a long-standing research problem for automatic speech recognition (ASR). A cross-utterance LM (CULM) is proposed in this paper, which augments the input to a standard long short-term memory (LSTM) LM with a context vector derived from past and future utterances using an extraction network. The extraction network uses another LSTM to encode surrounding utterances into vectors which are integrated into a context vector using either a projection of LSTM final hidden states, or a multi-head self-attentive layer. In addition, an acoustic error sampling technique is proposed to reduce the mismatch between training and test-time. This is achieved by considering possible ASR errors into the model training procedure, and can therefore improve the word error rate (WER). Experiments performed on both AMI and Switchboard datasets show that CULMs outperform the LSTM LM baseline WER. In particular, the CULM with a self-attentive layer-based extraction network and acoustic error sampling achieves 0.6% absolute WER reduction on AMI, 0.3% WER reduction on the Switchboard part and 0.9% WER reduction on the Callhome part of Eval2000 test set over the respective baselines.

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。

↑