A Comparison between Deep Neural Nets and Kernel Acoustic Models for Speech Recognition
This work addresses model selection in speech recognition for researchers, offering an incremental improvement by adapting existing techniques to enhance performance.
The paper compared kernel methods and deep neural networks (DNNs) for speech recognition, finding that kernel models match DNNs in perplexity and frame-level accuracy but DNNs are better at token-error-rates due to reducing perplexity and entropy. It proposed entropy regularized perplexity for model selection, which improved performance for both models and narrowed the gap, with results demonstrated on Broadcast News.
We study large-scale kernel methods for acoustic modeling and compare to DNNs on performance metrics related to both acoustic modeling and recognition. Measuring perplexity and frame-level classification accuracy, kernel-based acoustic models are as effective as their DNN counterparts. However, on token-error-rates DNN models can be significantly better. We have discovered that this might be attributed to DNN's unique strength in reducing both the perplexity and the entropy of the predicted posterior probabilities. Motivated by our findings, we propose a new technique, entropy regularized perplexity, for model selection. This technique can noticeably improve the recognition performance of both types of models, and reduces the gap between them. While effective on Broadcast News, this technique could be also applicable to other tasks.