Adaptive Multi-Corpora Language Model Training for Speech Recognition
This work addresses adaptation challenges in automatic speech recognition systems, offering an incremental improvement over static sampling methods.
The paper tackled the problem of improving neural network language model adaptation for speech recognition by dynamically adjusting sampling probabilities from multiple corpora during training, resulting in up to 7% and 9% relative word error rate reductions on in-domain and out-of-domain tasks.
Neural network language model (NNLM) plays an essential role in automatic speech recognition (ASR) systems, especially in adaptation tasks when text-only data is available. In practice, an NNLM is typically trained on a combination of data sampled from multiple corpora. Thus, the data sampling strategy is important to the adaptation performance. Most existing works focus on designing static sampling strategies. However, each corpus may show varying impacts at different NNLM training stages. In this paper, we introduce a novel adaptive multi-corpora training algorithm that dynamically learns and adjusts the sampling probability of each corpus along the training process. The algorithm is robust to corpora sizes and domain relevance. Compared with static sampling strategy baselines, the proposed approach yields remarkable improvement by achieving up to relative 7% and 9% word error rate (WER) reductions on in-domain and out-of-domain adaptation tasks, respectively.