Speech Model Pre-training for End-to-End Spoken Language Understanding
This work addresses data efficiency for end-to-end SLU systems, which is an incremental improvement for speech processing applications.
The paper tackles the challenge of high data requirements for end-to-end spoken language understanding (SLU) by proposing a pre-training method that predicts words and phonemes to learn better features, resulting in improved performance on a new dataset, Fluent Speech Commands, with gains in both full and small-data settings.
Whereas conventional spoken language understanding (SLU) systems map speech to text, and then text to intent, end-to-end SLU systems map speech directly to intent through a single trainable model. Achieving high accuracy with these end-to-end models without a large amount of training data is difficult. We propose a method to reduce the data requirements of end-to-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.