Whereas conventional spoken language understanding (SLU) systems map speech to text, and then text to intent, end-toend SLU systems map speech directly to intent through a single trainable model. Achieving high accuracy with these end-toend models without a large amount of training data is difficult. We propose a method to reduce the data requirements of endto-end SLU in which the model is first pre-trained to predict words and phonemes, thus learning good features for SLU. We introduce a new SLU dataset, Fluent Speech Commands, and show that our method improves performance both when the full dataset is used for training and when only a small subset is used. We also describe preliminary experiments to gauge the model's ability to generalize to new phrases not heard during training.

No abstract

This paper describes a novel method of live keyword spotting using a two-stage time delay neural network. The model is trained using transfer learning: initial training with phone targets from a large speech corpus is followed by training with keyword targets from a smaller data set. The accuracy of the system is evaluated on two separate tasks. The first is the freely available Google Speech Commands dataset. The second is an in-house task specifically developed for keyword spotting. The results show significant improvements in false accept and false reject rates in both clean and noisy environments when compared with previously known techniques. Furthermore, we investigate various techniques to reduce computation in terms of multiplications per second of audio. Compared to recently published work, the proposed system provides up to 89% savings on computational complexity.

No abstract

In this paper, we present a new method for recognizing tones in continuous speech for tonal languages. The method works by converting the speech signal to a cepstrogram, extracting a sequence of cepstral features using a convolutional neural network, and predicting the underlying sequence of tones using a connectionist temporal classification (CTC) network. The performance of the proposed method is evaluated on a freely available Mandarin Chinese speech corpus, AISHELL-1, and is shown to outperform the existing techniques in the literature in terms of tone error rate (TER).