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

TL;DR

This paper investigates how different layers of neural machine translation models encode linguistic information by evaluating their representations on part-of-speech and semantic tagging tasks, revealing layer-specific strengths.

Contribution

It introduces a method to analyze NMT representations across layers using classifier performance on linguistic tasks, providing new insights into what NMT models learn at each layer.

Findings

Higher layers encode semantic information better.

Lower layers are more effective for part-of-speech tagging.

Target language has little impact on source-side representations.

Abstract

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.