The provided empirical evidences show that CsaNMT sets a new level of performance among existing augmentation techniques, improving on the state-of-the-art by a large margin. We conduct extensive experiments on both rich-resource and low-resource settings involving various language pairs, including WMT14 English\rightarrowGerman,French, NIST Chinese\rightarrowEnglish and multiple low-resource IWSLT translation tasks. In this paper, we present a novel data augmentation paradigm termed Continuous Semantic Augmentation (CsaNMT), which augments each training instance with an adjacency semantic region that could cover adequate variants of literal expression under the same meaning. Although data augmentation is widely used to enrich the training data, conventional methods with discrete manipulations fail to generate diverse and faithful training samples. However, it is commonly observed that the generalization performance of the model is highly influenced by the amount of parallel data used in training. %X The principal task in supervised neural machine translation (NMT) is to learn to generate target sentences conditioned on the source inputs from a set of parallel sentence pairs, and thus produce a model capable of generalizing to unseen instances. %I Association for Computational Linguistics %S Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) %T Learning to Generalize to More: Continuous Semantic Augmentation for Neural Machine Translation The core codes are contained in Appendix E. The principal task in supervised neural machine translation (NMT) is to learn to generate target sentences conditioned on the source inputs from a set of parallel sentence pairs, and thus produce a model capable of generalizing to unseen instances. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)Īssociation for Computational Linguistics Learning to Generalize to More: Continuous Semantic Augmentation for Neural Machine Translation The core codes are contained in Appendix E.", We conduct extensive experiments on both rich-resource and low-resource settings involving various language pairs, including WMT14 English →, NIST Chinese$\rightarrow$English and multiple low-resource IWSLT translation tasks. Abstract The principal task in supervised neural machine translation (NMT) is to learn to generate target sentences conditioned on the source inputs from a set of parallel sentence pairs, and thus produce a model capable of generalizing to unseen instances.
0 Comments
Leave a Reply. |