In this work we explored building automatic speech recognition models for transcribing doctor patient conversation. We collected a large scale dataset of clinical conversations (14,000 hr), designed the task to represent the real word scenario, and explored several alignment approaches to iteratively improve data quality. We explored both CTC and LAS systems for building speech recognition models. The LAS was more resilient to noisy data and CTC required more data clean up. A detailed analysis is provided for understanding the performance for clinical tasks. Our analysis showed the speech recognition models performed well on important medical utterances, while errors occurred in causal conversations. Overall we believe the resulting models can provide reasonable quality in practice.
@article{arxiv.1711.07274,
title = {Speech recognition for medical conversations},
author = {Chung-Cheng Chiu and Anshuman Tripathi and Katherine Chou and Chris Co and Navdeep Jaitly and Diana Jaunzeikare and Anjuli Kannan and Patrick Nguyen and Hasim Sak and Ananth Sankar and Justin Tansuwan and Nathan Wan and Yonghui Wu and Xuedong Zhang},
journal= {arXiv preprint arXiv:1711.07274},
year = {2018}
}