What do Neural Machine Translation Models Learn about Morphology?

By Yonatan Belinkov, Nadir Durrani, Fahim Dalvi, Hassan Sajjad, James R. Glass. 2017

449 citations recorded by the source corpus.

Neural machine translation (MT) models obtain state-of-the-art performance while maintaining a simple, end-to-end architecture. However, little is known about what these models learn about source and target languages during the training process. In this work, we analyze the representations learned by neural MT models at various levels of granularity and empirically evaluate the quality of the representations for learning morphology through extrinsic part-of-speech and morphological tagging tasks. We conduct a thorough investigation along several parameters: word-based vs. character-based representations, depth of the encoding layer, the identity of the target language, and encoder vs. decoder representations. Our data-driven, quantitative evaluation sheds light on important aspects in the neural MT system and its ability to capture word structure.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Natural Language Processing

Citations in the corpus

Dissecting Contextual Word Embeddings: Architecture and Representation (2018, methodology): “Belinkov et al. (2017) used linear classiﬁers to determine whether neural machine translation systems learned morphology and POS tags.”
Deep Contextualized Word Representations (2018, methodology): “In an RNN-based encoder-decoder machine translation system, Belinkov et al. (2017) showed that the representations learned at the ﬁrst layer in a 2-layer LSTM encoder are better at predicting POS tags then second layer.”
The Bottom-up Evolution of Representations in the Transformer: A Study with Machine Translation and Language Modeling Objectives (2019, background): “…of linguistic analysis have been considered including morphology (Belinkov et al., 2017a; Dalvi et al., 2017; Bisazza and Tump, 2018), syntax (Shi et al., 2016; Tenney et al., 2019b) and semantics (Hill et al., 2017; Belinkov et al., 2017b; Raganato and Tiedemann, 2018; Tenney et al., 2019b).”
Learning General Purpose Distributed Sentence Representations via Large Scale Multi-task Learning (2018, methodology): “The addition of constituency parsing improves performance on sentence relatedness (SICK-R) and entailment (SICK-E) consistent with observations made by Bowman et al. (2016). In Table 4, we show that simply training an MLP on top of our fixed sentence representations outperforms several strong & comp”
BLOOM: A 176B-Parameter Open-Access Multilingual Language Model (2022, background): “While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public.”
How Can We Know What Language Models Know? (2019, background): “…by using extrinsic probing tasks to examine whether certain linguistic properties can be predicted from those representations (Shi et al., 2016; Linzen et al., 2016; Belinkov et al., 2017), or by ablations to the models to investigate how behavior varies (Li et al., 2016b; Smith et al., 2017).”
Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned (2019, background): “Different levels of linguistic analysis have been considered including morphology (Belinkov et al., 2017a; Dalvi et al., 2017; Bisazza and Tump, 2018), syntax (Shi et al., 2016) and semantics (Hill et al., 2017; Belinkov et al., 2017b; Raganato and Tiedemann, 2018).”
Locating and Editing Factual Associations in GPT (2022, background): “…are encoded in internal model representations, most commonly by training a probing classiﬁer to predict said properties from the representations (Ettinger et al., 2016; Adi et al., 2017; Hupkes et al., 2018; Conneau et al., 2018; Belinkov et al., 2017; Belinkov & Glass, 2019, inter alia).”
Fast Model Editing at Scale (2021, background): “…locus of various types of knowledge in language models, using learned probe models or iterative search procedures to test for linguistic structures (Belinkov et al., 2017; Conneau et al., 2018; Hewitt and Manning, 2019) 2020), which assume access to many datapoints representing the knowledge to be…”
Factual Probing Is [MASK]: Learning vs. Learning to Recall (2021, background): “…has been explored recently in linguistic probing, which aims to explore the linguistic properties encoded in contextualized word representations (Belinkov et al., 2017; Tenney et al., 2019; Lin et al., 2019)—for example, by seeing if a classifier can predict that “chef ” is the nominal subject…”
Probing Classifiers: Promises, Shortcomings, and Advances (2021, background): “7 “evaluate the quality of the trained classiﬁer on the given task as a proxy to the quality of the extracted representations” (Belinkov et al. 2017a).”
Measuring and Improving Consistency in Pretrained Language Models (2021, background): “Many papers have found neural models (especially PLMs) to naturally encode syntax (Linzen et al., 2016; Belinkov et al., 2017; Marvin and Linzen, 2018; Belinkov and Glass, 2019; Goldberg, 2019; Hewitt and Manning, 2019).”
What Does BERT Look at? An Analysis of BERT’s Attention (2019, background): “Recent work has investigated this question by examining the outputs of language models on carefully chosen input sentences (Linzen et al., 2016) or examining the internal vector representations of the model through methods such as probing classiﬁers (Adi et al., 2017; Belinkov et al., 2017).”
What do you learn from context? Probing for sentence structure in contextualized word representations (2019, methodology): “Recent work has explored many token-level properties of these representations, such as their ability to capture part-of-speech tags (Blevins et al., 2018; Belinkov et al., 2017b; Shi et al., 2016), morphology (Belinkov et al., 2017a;b), or word-sense disambiguation (Peters et al., 2018a).”
Synthetic and Natural Noise Both Break Neural Machine Translation (2017, background): “In many cases, these models actually discover an impressive amount of morphological information about a language (Belinkov et al., 2017a).”
Designing and Interpreting Probes with Control Tasks (2019, methodology): “, 2019), but multi-layer perceptron (MLP) probes are at least as popular (Belinkov et al., 2017; Conneau et al., 2018; Adi et al., 2017; Tenney et al., 2019; Ettinger et al., 2018).”
A Structural Probe for Finding Syntax in Word Representations (2019, methodology): “To this end, probing methods are designed to evaluate the extent to which representations of language encode particular knowledge of interest, like part-of-speech (Belinkov et al., 2017), morphology (Peters et al., 2018a), or sentence length (Adi et al., 2017).”