Fuzzy Alignments in Directed Acyclic Graph for Non-Autoregressive Machine Translation
This addresses performance degradation in non-autoregressive translation, which is crucial for reducing decoding latency in real-time applications, though it appears incremental as it builds on existing graph-based methods.
The paper tackles the multi-modality problem in non-autoregressive machine translation by proposing fuzzy alignments in directed acyclic graphs, which improves translation performance and sets a new state of the art on major WMT benchmarks.
Non-autoregressive translation (NAT) reduces the decoding latency but suffers from performance degradation due to the multi-modality problem. Recently, the structure of directed acyclic graph has achieved great success in NAT, which tackles the multi-modality problem by introducing dependency between vertices. However, training it with negative log-likelihood loss implicitly requires a strict alignment between reference tokens and vertices, weakening its ability to handle multiple translation modalities. In this paper, we hold the view that all paths in the graph are fuzzily aligned with the reference sentence. We do not require the exact alignment but train the model to maximize a fuzzy alignment score between the graph and reference, which takes captured translations in all modalities into account. Extensive experiments on major WMT benchmarks show that our method substantially improves translation performance and increases prediction confidence, setting a new state of the art for NAT on the raw training data.