Word Order Does Not Matter For Speech Recognition

10/12/2021

∙

by   Vineel Pratap, et al.

∙

0

∙

share

In this paper, we study training of automatic speech recognition system in a weakly supervised setting where the order of words in transcript labels of the audio training data is not known. We train a word-level acoustic model which aggregates the distribution of all output frames using LogSumExp operation and uses a cross-entropy loss to match with the ground-truth words distribution. Using the pseudo-labels generated from this model on the training set, we then train a letter-based acoustic model using Connectionist Temporal Classification loss. Our system achieves 2.3 LibriSpeech, which closely matches with the supervised baseline's performance.