областях • Задачи перевода и поиска имеют собственные метрики качества [1][2][3] • Нет зависимости между тем высоким WER и правильным пониманием [4] • Итоговая оценка качества производится человеком [1] X. He, L. Deng and A. Acero, "Why word error rate is not a good metric for speech recognizer training for the speech translation task?," 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Prague, 2011, pp. 5632-5635, doi: 10.1109/ICASSP.2011.5947637. [2] J. Garofolo, C. Auzanne, and E. Voorhees, “The TREC spoken document retrieval track: A success story,” in Proceedings of the Eighth Text REtrieval Conference (TREC 8), 1999. [3] Grangier, D., Vinciarelli, A., and Bourlard, H., “Information retrieval on noisy text,” IDIAP-COM 03-08, IDIAP, 2003. [4] Y. Wang, A. Acera, and C. Chelba, “Is word error rate a good indicator for spoken language understanding accuracy,” in IEEE Workshop on Automatic Speech Recognition and Understanding. IEEE, 2003, pp. 577–582.