Preprint
Does Joint Training Really Help Cascaded Speech Translation?
Viet Anh Khoa Tran et al., arXiv, 2022
Read the originalRead on October 17, 2026
- What it is
- A 2022 study by Tran, Thulke, Gao, Herold and Ney that compares cascaded speech translation, which runs a speech recognizer and then a translation model, with joint training methods on the MuST-C English to German benchmark. Its abstract names error propagation from the recognizer as the main weakness of cascades.
- How much weight it carries
- A preprint on arXiv, not a peer-reviewed venue at the time of reading. Its experiments use one benchmark and one language pair, English to German, and a 2022 recognizer, so the size of the effect in modern apps is unknown.
Facts taken from this source
- The authors of a 2022 study call error propagation from the speech recognition system a fundamental challenge for cascaded speech translation that still remains.
“fundamental challenges such as error propagation from the automatic speech recognition system still remain.” (Abstract)