On supervised learning from sequential data with applications for speech regognition

By Mike Schuster. 1999

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Yu Zhang
2. Xu Tan
3. Ye Jia
4. Ron J. Weiss
5. Heiga Zen
6. Wei-Ning Hsu
7. Jinyu Li
8. Tao Qin

Citations in the corpus

• Deep learning in neural networks: An overview (2014, background): “orithms able to deal with long time lags (Sch¨afer et al., 2006; Martens and Sutskever, 2011; Zimmermann et al., 2012; Koutn´ık et al., 2014). Bi-directional RNNs (BRNNs) (Schuster and Paliwal, 1997; Schuster, 1999) are designed for input sequences whose starts and ends are known in advance, such as”
• Deep mixture density networks for acoustic modeling in statistical parametric speech synthesis (2014, background): “It is also known that the outputs of an ANN trained by minimizing the squared loss function approximates the conditional mean of the outputs in the training data [17,18].”
• Unidirectional long short-term memory recurrent neural network with recurrent output layer for low-latency speech synthesis (2015, methodology): “To use both past and future inputs for prediction, Schuster proposed the bidirectional RNN architecture [15].”
• Framewise phoneme classification with bidirectional LSTM and other neural network architectures (2005, unknown): “For the MLP, RNN and LSTM nets we give the best results, and those achieved with least contextual .9% 74.2% - RNN (Robinson, 1994; Schuster, 1999) 70.6% 65.3% - BRNN (Schuster, 1999) 72.1% 65.1% information (i.e. with no target delay / time-window).”
• Supervised Sequence Labelling with Recurrent Neural Networks (2012, background): “, 2001; Chen and Chaudhari, 2004) and speech processing (Schuster, 1999; Fukada et al., 1999).”
• Framewise phoneme classification with bidirectional LSTM networks (2005, unknown): “Bidirectional recurrent neural nets (BRNNs) have given significantly improved results in sequence learning tasks, notably protein structure prediction (PSP) [1], [6] and speech processing [21], [9].”
• Natural TTS Synthesis by Conditioning Wavenet on MEL Spectrogram Predictions (2017, methodology): “We also experimented with a log-likelihood loss by modeling the output distribution with a Mixture Density Network [23, 24] to avoid assuming a constant variance over time, but found that these were more difficult to train and they did not lead to better sounding samples.”