Lead2Gold: Towards exploiting the full potential of noisy transcriptions for speech recognition
This addresses the challenge of leveraging imperfect training data for speech recognition, but it is incremental as it builds on existing sequence-to-sequence frameworks.
The paper tackles the problem of training automatic speech recognition systems with noisy transcriptions by introducing Lead2Gold, a method that uses a differentiable beam search with a noise model to find better transcriptions, resulting in improved ASR accuracy over a baseline that ignores transcription noise.
The transcriptions used to train an Automatic Speech Recognition (ASR) system may contain errors. Usually, either a quality control stage discards transcriptions with too many errors, or the noisy transcriptions are used as is. We introduce Lead2Gold, a method to train an ASR system that exploits the full potential of noisy transcriptions. Based on a noise model of transcription errors, Lead2Gold searches for better transcriptions of the training data with a beam search that takes this noise model into account. The beam search is differentiable and does not require a forced alignment step, thus the whole system is trained end-to-end. Lead2Gold can be viewed as a new loss function that can be used on top of any sequence-to-sequence deep neural network. We conduct proof-of-concept experiments on noisy transcriptions generated from letter corruptions with different noise levels. We show that Lead2Gold obtains a better ASR accuracy than a competitive baseline which does not account for the (artificially-introduced) transcription noise.