Non-Uniform MCE Training of Deep Long Short-Term Memory Recurrent Neural Networks for Keyword Spotting

Abstract

It has been shown in [1,2] that improved performance can be achieved by formulating the keyword spotting as a non-uniform error automatic speech recognition problem. In this work, we discriminatively train a deep bidirectional long short-term memory (BLSTM) -hidden Markov model (HMM) based acoustic model with non-uniform boosted minimum classification error (BMCE) criterion which imposes more significant error cost on the keywords than those on the non-keywords. By introducing the BLSTM, the context information in both the past and the future are stored and updated to predict the desired output and the long-term dependencies within the speech signal are well captured. With non-uniform BMCE objective, the BLSTM is trained so that the recognition errors related to the keywords are remarkably reduced. The BLSTM is optimized using backpropagation through time and stochastic gradient descent. The keyword spotting system is implemented within weighted finite state transducer framework. The proposed method achieves 5.49% and 7.37% absolute figure-of-merit improvements respectively over the BLSTM and the feedforward deep neural network baseline systems trained with cross-entropy criterion for the keyword spotting task on Switchboard-1 Release 2 dataset.

The Multi-Domain International Search on Speech 2020 ALBAYZIN Evaluation: Overview, Systems, Results, Discussion and Post-Evaluation Analyses

“…Spoken term detection has been a hot topic in the past few years due to its many applications and has received a great deal of interest from many outstanding companies and research institutes, such as IBM [14,[32][33][34][35][36], BBN [37][38][39], SRI & OGI [40][41][42], BUT [17,43,44], Microsoft [45], QUT [46,47], JHU [16,[48][49][50], Fraunhofer IAIS/NTNU/TUD [15], NTU [31,51], IDIAP [52] and Google [21], among others. Within an STD system, the ASR subsystem uses mostly word-based speech recognition [24,41,[53][54][55][56][57][58][59] due to its better performance in comparison with subword-based approaches.…”

ALBAYZIN 2018 spoken term detection evaluation: a multi-domain international evaluation in Spanish

“…For the ASR stage, word-based speech recognition has been widely used [35,[48][49][50][51][52][53][54], since this typically yields better performance than subword-based ASR [55][56][57][58][59][60][61][62] due to the lexical and language model (LM) information employed by the word-based ASR. However, one of the main drawbacks of word-based ASR is that it can only detect in-vocabulary (INV) terms.…”