Framewise phoneme classification with bidirectional LSTM networks

Framewise phoneme classification with bidirectional LSTM networks
复制标题

DOI：
10.1109/ijcnn.2005.1556215
复制
发表时间：
2005-12
期刊：
Proceedings. 2005 IEEE International Joint Conference on Neural Networks, 2005.
影响因子：
--
通讯作者：
Alex Graves;J. Schmidhuber
Alex Graves;J. Schmidhuber
中科院分区：
其他
文献类型：
--
作者：
Alex Graves;J. Schmidhuber

文献摘要

被引文献

相似文献

在本文中，我们首次将双向训练应用于长短期记忆（LSTM）网络。我们还提出了LSTM学习算法的修改后的完整梯度版本。我们讨论了连续语音识别的框架音素分类的意义，以及使用双向网络在线因果任务的有效性。在TIMIT语音数据库上，我们测量了LSTM和传统递归神经网络（RNN）的双向和单向变体的帧级音素分类得分。我们发现双向LSTM优于RNN和单向LSTM。
In this paper, we apply bidirectional training to a long short term memory (LSTM) network for the first time. We also present a modified, full gradient version of the LSTM learning algorithm. We discuss the significance of framewise phoneme classification to continuous speech recognition, and the validity of using bidirectional networks for online causal tasks. On the TIMIT speech database, we measure the framewise phoneme classification scores of bidirectional and unidirectional variants of both LSTM and conventional recurrent neural networks (RNNs). We find that bidirectional LSTM outperforms both RNNs and unidirectional LSTM.