Samsung Research Philippines - Datasaur AI’s Submission for the WMT22 Large Scale Multilingual Translation Task
Samsung Research Philippines - Datasaur AI’s Submission for the WMT22 Large Scale Multilingual Translation Task
复制标题
三星菲律宾研究院 - Datasaur AI 提交的 WMT22 大规模多语言翻译任务
DOI：
--
复制
发表时间：
2022
期刊：
影响因子：
--
通讯作者：
Lintang Sutawika
中科院分区：
文献类型：
--
作者：
Jan Christian Blaise Cruz;Lintang Sutawika
本文介绍了三星菲律宾研究院- Datasaur AI联合团队为WMT 22大规模多语言非洲翻译共享任务提交的报告。我们将竞赛作为一种探索任务组合作为低资源多语言翻译解决方案的方式，使用适配器融合来联合收割机组合多个任务适配器，这些任务适配器学习总翻译对的子集。我们的最终模型显示，与同时在多个方向上训练的单个模型系统相比，我们参与的44个翻译方向中有32个方向的性能有所提高。
This paper describes the submission of the joint Samsung Research Philippines - Datasaur AI team for the WMT22 Large Scale Multilingual African Translation shared task. We approach the contest as a way to explore task composition as a solution for low-resource multilingual translation, using adapter fusion to combine multiple task adapters that learn subsets of the total translation pairs. Our final model shows performance improvements in 32 out of the 44 translation directions that we participate in when compared to a single model system trained on multiple directions at once.