Modeling Dual Read/Write Paths for Simultaneous Machine Translation
This work addresses a key bottleneck in simultaneous machine translation for real-time applications, offering an incremental improvement over existing methods.
The paper tackles the lack of direct supervision for read/write paths in simultaneous machine translation by proposing a dual-path method with duality constraints, which outperforms strong baselines on En-Vi and De-En tasks across all latency levels.
Simultaneous machine translation (SiMT) outputs translation while reading source sentence and hence requires a policy to decide whether to wait for the next source word (READ) or generate a target word (WRITE), the actions of which form a read/write path. Although the read/write path is essential to SiMT performance, no direct supervision is given to the path in the existing methods. In this paper, we propose a method of dual-path SiMT which introduces duality constraints to direct the read/write path. According to duality constraints, the read/write path in source-to-target and target-to-source SiMT models can be mapped to each other. As a result, the two SiMT models can be optimized jointly by forcing their read/write paths to satisfy the mapping. Experiments on En-Vi and De-En tasks show that our method can outperform strong baselines under all latency.