A Structural Probe for Finding Syntax in Word Representations

By John Hewitt, Christopher D. Manning. 2019

1,402 citations recorded by the source corpus.

Recent work has improved our ability to detect linguistic knowledge in word representations. However, current methods for detecting syntactic knowledge do not test whether syntax trees are represented in their entirety. In this work, we propose a structural probe, which evaluates whether syntax trees are embedded in a linear transformation of a neural network’s word representation space. The probe identifies a linear transformation under which squared L2 distance encodes the distance between words in the parse tree, and one in which squared L2 norm encodes depth in the parse tree. Using our probe, we show that such transformations exist for both ELMo and BERT but not in baselines, providing evidence that entire syntax trees are embedded implicitly in deep models’ vector geometry.

Subfields: Artificial Intelligence, Natural Language Processing, Representation & Self-Supervised Learning

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 Natural Language Processing

In Representation & Self-Supervised Learning

Citations in the corpus

How Context Affects Language Models' Factual Predictions (2020, background): “A variety of “probes” have been developed to analyse the syntactic structures, such as syntax trees [Marvin and Linzen, 2018, Hewitt and Manning, 2019, Vig and Belinkov, 2019], negative polarity items [Warstadt and Bowman, 2019, Warstadt et al., 2019], semantic fragments [Richardson et al., 2019],…”
oLMpics-On What Language Model Pre-training Captures (2019, background): “…that these LM compute, revealing that they encode substantial amounts of syntax and semantics (Linzen et al., 2016b; Peters et al., 2018b; Tenney et al., 2019b; Goldberg, 2019; Hewitt and Manning, 2019; Ten-ney et al., 2019a; Shwartz and Dagan, 2019; Lin et al., 2019; Coenen et al., 2019).”
Injecting Numerical Reasoning Skills into Language Models (2020, background): “Recently, models trained on large amounts of data with a language modeling (LM) objective, have shown great promise in natural language processing, exhibiting surprising amounts of knowledge and information (Peters et al., 2018; Devlin et al., 2019; Liu et al., 2019; Lan et al., 2019; Petroni et al.”
Enhancing the Transformer with Explicit Relational Encoding for Math Problem Solving (2019, methodology): “This was shown for BERT’s hidden activation states (Hewitt & Manning, 2019; Tenney et al., 2019) and, most directly related to the present work, for the graph implicit in BERT’s attention weights (Coenen et al., 2019; Lin et al., 2019).”
Fast Model Editing at Scale (2021, methodology): “In the NLP literature, many papers have investigated the 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) or facts about the worl”
Causal Analysis of Syntactic Agreement Mechanisms in Neural Language Models (2021, background): “, 2019b) or full dependency parses (Hewitt and Manning, 2019; Chi et al., 2020) as a proxy for discovering latent syntactic knowledge within the model.”
Measuring and Improving Consistency in Pretrained Language Models (2021, background): “, 2020) perform on linguistic tasks (Goldberg, 2019; Hewitt and Manning, 2019; Tenney et al., 2019; Elazar et al., 2020), commonsense (Forbes et al.”
Amnesic Probing: Behavioral Explanation with Amnesic Counterfactuals (2020, background): “, 2019), syntax (Tenney et al., 2019b; Lin et al., 2019; Reif et al., 2019; Hewitt and Manning, 2019; Liu et al., 2019a), world knowledge (Petroni et al.”
Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data (2020, background): “4 Hewitt and Manning (2019) ﬁnd information analogous to unlabeled dependency structures in the word vectors provided by ELMo and BERT (trained on English).”
Designing and Interpreting Probes with Control Tasks (2019, methodology): “We also examine the representations of an un-trained BiLSTM run on the non-contextual character CNN word embeddings of ELMo, shown to be a strong baseline contextualization method, but without any linguistic knowledge learned from context (Zhang and Bowman, 2018; Hewitt and Manning, 2019).”
A Primer in BERTology: What We Know About How BERT Works (2020, background): “Hewitt and Manning (2019) had the most success reconstructing syntactic tree depth from the middle BERT layers (6-9 for base-BERT, 14-19 for BERT-large).”
Revisiting Few-sample BERT Fine-tuning (2020, methodology): “[16] John Hewitt and Christopher D.”
Biological structure and function emerge from scaling unsupervised learning to 250 million protein sequences (2019, methodology): “…To learn the distance projection used in the phylogenetic organization analysis, we applied an adapted version of the technique of Hewitt & Manning (2019), originally employed toward learning a natural language parse-tree encoding; we modify the technique to instead learn a phylogenetic…”
BERTology Meets Biology: Interpreting Attention in Protein Language Models (2020, background): “Interpreting attention on natural language sequences is a well-established area of research [12, 30, 87, 90].”
Pre-Trained Models: Past, Present and Future (2021, neutral): “They regard neural networks as parameterized functions and use knowledge distillation (Hinton et al., 2014) to import and export modeledge.”
Open Sesame: Getting inside BERT’s Linguistic Knowledge (2019, background): “C L ] 4 J un 2 01 9 capture hierarchy-sensitive, syntactic dependencies, and even support the extraction of dependency parses (Hewitt and Manning, 2019).”
Analyzing the Structure of Attention in a Transformer Language Model (2019, background): “Recent work suggests that the Transformer implicitly encodes syntactic information such as dependency parse trees (Hewitt and Manning, 2019; Raganato and Tiedemann, 2018), anaphora (Voita et al.”