Accelerating recurrent neural network language model based online speech recognition system
This work addresses efficiency bottlenecks for real-time speech recognition systems, though it is incremental as it builds on existing RNNLM methods.
The paper tackles the problem of slow recurrent neural network language models (RNNLMs) in online speech recognition by introducing lossy compression of history vectors and CPU-GPU hybrid computation, achieving a 1.23x speed improvement with minimal accuracy loss and enabling real-time recognition with a 4x speedup.
This paper presents methods to accelerate recurrent neural network based language models (RNNLMs) for online speech recognition systems. Firstly, a lossy compression of the past hidden layer outputs (history vector) with caching is introduced in order to reduce the number of LM queries. Next, RNNLM computations are deployed in a CPU-GPU hybrid manner, which computes each layer of the model on a more advantageous platform. The added overhead by data exchanges between CPU and GPU is compensated through a frame-wise batching strategy. The performance of the proposed methods evaluated on LibriSpeech test sets indicates that the reduction in history vector precision improves the average recognition speed by 1.23 times with minimum degradation in accuracy. On the other hand, the CPU-GPU hybrid parallelization enables RNNLM based real-time recognition with a four times improvement in speed.