CUNI Non-Autoregressive System for the WMT 22 Efficient Translation Shared Task
This work provides a baseline for efficient translation research, but it is incremental as it builds on existing methods without introducing major innovations.
The paper tackled the problem of efficient machine translation by developing a non-autoregressive system for the WMT 22 Efficient Translation Shared Task, achieving results used to compare non-autoregressive and autoregressive models with a focus on decoding speed evaluation.
We present a non-autoregressive system submission to the WMT 22 Efficient Translation Shared Task. Our system was used by Helcl et al. (2022) in an attempt to provide fair comparison between non-autoregressive and autoregressive models. This submission is an effort to establish solid baselines along with sound evaluation methodology, particularly in terms of measuring the decoding speed. The model itself is a 12-layer Transformer model trained with connectionist temporal classification on knowledge-distilled dataset by a strong autoregressive teacher model.