通过多模型集成提升无监督机器翻译效果
Ensemble Self-Training for Unsupervised Machine Translation
- 多个模型共享任务但使用不同辅助语言，实现结构化多样性
- 用双向预测平均生成伪平行数据，平均提升1.7 chrF（英译其他）
- 部署时选最优模型，保持单模型推理效率
我们提出一种基于集成的自训练框架，用于无监督神经机器翻译（UNMT）。从主语言对出发，训练多个共享翻译任务但使用不同辅助语言的UNMT模型，从而在模型间引入结构化多样性。随后，通过词级集成解码，对两个方向的模型预测进行平均，生成主语言对的伪翻译数据。这些集成生成的合成平行数据被用于进一步训练各模型，使模型在共享监督下持续优化。部署时，依据验证性能选择单一模型，保持单模型推理开销。实验表明，该方法在统计上显著优于单模型基线，英译其他语言平均提升1.7 chrF，其他语言译英平均提升0.67 chrF。
原文摘要 · Abstract (English)
We present an ensemble-driven self-training framework for unsupervised neural machine translation (UNMT). Starting from a primary language pair, we train multiple UNMT models that share the same translation task but differ in an auxiliary language, inducing structured diversity across models. We then generate pseudo-translations for the primary pair using token-level ensemble decoding, averaging model predictions in both directions. These ensemble outputs are used as synthetic parallel data to further train each model, allowing the models to improve via shared supervision. At deployment time, we select a single model by validation performance, preserving single-model inference cost. Experiments show statistically significant improvements over single-model UNMT baselines, with mean gains of 1.7 chrF when translating from English and 0.67 chrF when translating into English.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。