RNNs for audio transcripts processing
Speech recognition using RNN. Although transformer is popular with speech recognition, other model strustures have their advantages especially in deployment in production. This is a sample of RNN model for speech recognition, with word error count of 16%. With more data and further fine-tuning, it has potential to on par with Transformer models.