On Addressing Practical Challenges for RNN-Transducer

TL;DR

This paper presents practical solutions for deploying RNN-Transducer speech recognition systems, including domain adaptation without audio data, word-level timestamping, and confidence scoring, validated on Microsoft data.

Contribution

It introduces a splicing data method for domain adaptation, a shared encoder for precise timestamping, and confidence scoring techniques with minimal computational overhead.

Findings

Splicing data method reduces word error rate by 58.03%.

Time stamping achieves less than 50ms error on average.

High confidence annotation with low computational cost.

Abstract

In this paper, several works are proposed to address practical challenges for deploying RNN Transducer (RNN-T) based speech recognition system. These challenges are adapting a well-trained RNN-T model to a new domain without collecting the audio data, obtaining time stamps and confidence scores at word level. The first challenge is solved with a splicing data method which concatenates the speech segments extracted from the source domain data. To get the time stamp, a phone prediction branch is added to the RNN-T model by sharing the encoder for the purpose of force alignment. Finally, we obtain word-level confidence scores by utilizing several types of features calculated during decoding and from confusion network. Evaluated with Microsoft production data, the splicing data adaptation method improves the baseline and adaptation with the text to speech method by 58.03% and 15.25% relative word error rate reduction, respectively. The proposed time stamping method can get less than 50ms word timing difference from the ground truth alignment on average while maintaining the recognition accuracy of the RNN-T model. We also obtain high confidence annotation performance with limited computation cost.