Evaluating Layers of Representation in Neural Machine Translation on Part-of-Speech and Semantic Tagging Tasks
This provides insights into representation learning in NMT models, which is incremental for researchers in machine translation and natural language processing.
The paper investigates what neural machine translation (NMT) models learn about language by evaluating representations from different encoder layers on part-of-speech and semantic tagging tasks, finding that higher layers are better for semantics and lower layers for part-of-speech tagging, with little effect from the target language.
While neural machine translation (NMT) models provide improved translation quality in an elegant, end-to-end framework, it is less clear what they learn about language. Recent work has started evaluating the quality of vector representations learned by NMT models on morphological and syntactic tasks. In this paper, we investigate the representations learned at different layers of NMT encoders. We train NMT systems on parallel data and use the trained models to extract features for training a classifier on two tasks: part-of-speech and semantic tagging. We then measure the performance of the classifier as a proxy to the quality of the original NMT model for the given task. Our quantitative analysis yields interesting insights regarding representation learning in NMT models. For instance, we find that higher layers are better at learning semantics while lower layers tend to be better for part-of-speech tagging. We also observe little effect of the target language on source-side representations, especially with higher quality NMT models.