English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

While neural machine translation (NMT) models provide improved translationquality in an elegant, end-to-end framework, it is less clear what they learnabout language. Recent work has started evaluating the quality of vectorrepresentations learned by NMT models on morphological and syntactic tasks. Inthis paper, we investigate the representations learned at different layers ofNMT encoders. We train NMT systems on parallel data and use the trained modelsto extract features for training a classifier on two tasks: part-of-speech andsemantic tagging. We then measure the performance of the classifier as a proxyto the quality of the original NMT model for the given task. Our quantitativeanalysis yields interesting insights regarding representation learning in NMTmodels. For instance, we find that higher layers are better at learningsemantics while lower layers tend to be better for part-of-speech tagging. Wealso observe little effect of the target language on source-siderepresentations, especially with higher quality NMT models.

Evaluating Layers of Representation in Neural Machine Translation on Part-of-Speech and Semantic Tagging Tasks