Why word error rate is not a good metric for speech recognizer training for the speech translation task?

Xiaodong He, Li Deng, Alex Acero. Why word error rate is not a good metric for speech recognizer training for the speech translation task?. In Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2011, May 22-27, 2011, Prague Congress Center, Prague, Czech Republic. pages 5632-5635, IEEE, 2011. [doi]

@inproceedings{HeDA11,
  title = {Why word error rate is not a good metric for speech recognizer training for the speech translation task?},
  author = {Xiaodong He and Li Deng and Alex Acero},
  year = {2011},
  doi = {10.1109/ICASSP.2011.5947637},
  url = {http://dx.doi.org/10.1109/ICASSP.2011.5947637},
  tags = {translation},
  researchr = {https://researchr.org/publication/HeDA11},
  cites = {0},
  citedby = {0},
  pages = {5632-5635},
  booktitle = {Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2011, May 22-27, 2011, Prague Congress Center, Prague, Czech Republic},
  publisher = {IEEE},
  isbn = {978-1-4577-0539-7},
}