No abstract

No abstract

We present a simple yet powerful data augmentation method for boosting Neural Machine Translation (NMT) performance by leveraging information retrieved from a Translation Memory (TM). We propose and test two methods for augmenting NMT training data with fuzzy TM matches. Tests on the DGT-TM data set for two language pairs show consistent and substantial improvements over a range of baseline systems. The results suggest that this method is promising for any translation environment in which a sizeable TM is available and a certain amount of repetition across translations is to be expected, especially considering its ease of implementation.

This cross-sectional study investigated the impact of length of instruction, out-of-school exposure to foreign language input, and gender on learners' receptive vocabulary knowledge in two foreign languages: French (first foreign language) and English (second foreign language). The findings suggest that, although length of instruction correlated positively with vocabulary knowledge in English and French, the gains remained modest when out-of-school exposure to the foreign language input was limited. Despite fewer years of English instruction, participants' vocabulary knowledge in English was considerably larger than their French vocabulary knowledge, which can be explained by their large amounts of out-of-school exposure to English language input. Participants' online activities in particular had a positive effect on their vocabulary knowledge in English. Although gender influenced participants' engagement with online activities in English, gender did not have a direct effect on their vocabulary knowledge, as the structural equation modeling analysis showed.We would like to thank the four anonymous reviewers and Associate Editor Scott Crossley for their insightful comments on previous versions of this manuscript. We are also grateful to Britta Kestemont for her help in the data collection.