Abstract
In recent years, pretrained word embeddings have proved useful for multimodal neural machine translation (NMT) models to address the shortage of available datasets. However, the integration of pretrained word embeddings has not yet been explored extensively. Further, pretrained word embeddings in high dimensional spaces have been reported to suffer from the hubness problem. Although some debiasing techniques have been proposed to address this problem for other natural language processing tasks, they have seldom been studied for multimodal NMT models. In this study, we examine various kinds of word embeddings and introduce two debiasing techniques for three multimodal NMT models and two language pairs -- English-German translation and English-French translation. With our optimal settings, the overall performance of multimodal models was improved by up to +1.93 BLEU and +2.02 METEOR for English-German translation and +1.73 BLEU and +0.95 METEOR for English-French translation.
Abstract (translated)
近年来，预先训练的字嵌入已被证明是有用的多模态神经机器翻译（NMT）模型，以解决现有数据集的不足。然而，预训练词嵌入的整合还没有得到广泛的探索。此外，据报道，在高维空间中预训练的单词嵌入会产生Hubness问题。尽管有人提出了一些借记技术来解决其他自然语言处理任务中的这个问题，但对于多模NMT模型很少研究这些技术。在本研究中，我们研究了各种类型的单词嵌入，并介绍了三种多模态NMT模型和两种语言对的两种借记技术——英语-德语翻译和英语-法语翻译。在我们的优化设置下，多模态模型的整体性能提高了高达+1.93 Bleu和+2.02 Meteor（英语-德语翻译）和+1.73 Bleu和+0.95 Meteor（英语-法语翻译）。
URL
https://arxiv.org/abs/1905.10464
