The Volctrans Machine Translation System for WMT20
This work addresses machine translation for news domains, but it is incremental as it builds on existing Transformer methods with standard optimizations.
The paper tackled the WMT20 shared news translation task by developing the VolcTrans system, which achieved competitive results across 8 translation directions using Transformer-based models enhanced with techniques like data selection and model ensemble.
This paper describes our VolcTrans system on WMT20 shared news translation task. We participated in 8 translation directions. Our basic systems are based on Transformer, with several variants (wider or deeper Transformers, dynamic convolutions). The final system includes text pre-process, data selection, synthetic data generation, advanced model ensemble, and multilingual pre-training.