Assessing Composition in Sentence Vector Representations

By Allyson Ettinger, Ahmed Elgohary, Colin Phillips, Philip Resnik. 2018

84 citations recorded by the source corpus.

An important component of achieving language understanding is mastering the composition of sentence meaning, but an immediate challenge to solving this problem is the opacity of sentence vector representations produced by current neural sentence composition models. We present a method to address this challenge, developing tasks that directly target compositional meaning information in sentence vector representations with a high degree of precision and control. To enable the creation of these controlled tasks, we introduce a specialized sentence generation system that produces large, annotated sentence sets meeting specified syntactic, semantic and lexical constraints. We describe the details of the method and generation system, and then present results of experiments applying our method to probe for compositional information in embeddings from a number of existing sentence composition models. We find that the method is able to extract useful information about the differing capacities of these models, and we discuss the implications of our results with respect to these systems’ capturing of sentence information. We make available for public use the datasets used for these experiments, as well as the generation system.

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

Neural Network Acceptability Judgments (2018, methodology): “Examples of such tasks are to determine whether the sentence is in active or passive voice (Shi et al., 2016), whether the subject is singular or plural (Conneau et al., 2018), or whether a given token is under the scope of negation (Ettinger et al., 2018).”
Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference (2019, background): “To achieve this, we structured our vocabulary around Ettinger et al. (2018), who chose a vocabulary such that any noun can plausibly be the subject of any verb or the object of any transitive verb.”
Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data (2020, methodology): “The methodology of probing tasks (e.g. Adi et al., 2017; Ettinger et al., 2018) has been used to show that large LMs learn at least some information about phenomena such as English subject-verb agreement (Goldberg, 2019; Jawahar et al., 2019), constituent types, dependency labels, NER, and (core)…”
What do you learn from context? Probing for sentence structure in contextualized word representations (2019, background): “Often, probing tasks take the form of “challenge sets”, or test sets which are generated using templates and/or perturbations of existing test sets in order to isolate particular linguistic phenomena, e.g. compositional reasoning (Dasgupta et al., 2018; Ettinger et al., 2018).”
What BERT Is Not: Lessons from a New Suite of Psycholinguistic Diagnostics for Language Models (2019, neutral): “Some approaches use fine-grained classification tasks to probe information in sentence embeddings (Adi et al., 2016; Conneau et al., 2018; Ettinger et al., 2018), or token-level and other sub-sentence level information in contextual embeddings (Tenney et al.”
Designing and Interpreting Probes with Control Tasks (2019, background): “, 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).”
Targeted Syntactic Evaluation of Language Models (2018, background): “For a review, see Giannakidou (2011). confounds that may complicate the interpretation of the experiments (Ettinger et al., 2018).”
PAWS: Paraphrase Adversaries from Word Scrambling (2019, background): “This relates to prior work on probing sentence representations for their linguistic properties, such as how much syntactic information is encoded in representations (Conneau et al., 2018; Tenney et 2019; Ettinger et al., 2018).”
Investigating BERT’s Knowledge of Language: Five Analysis Methods with NPIs (2019, methodology): “4 Licensing Features We create a controlled set of 136,000 English sentences using an automated sentence generation procedure, inspired in large part by previous work by Ettinger et al. (2016, 2018), Marvin and Linzen (2018), Dasgupta et al. (2018), and Kann et al. (2019).”
Inherent Disagreements in Human Textual Inferences (2019, methodology): “…be useful in applications, and currently, NLI is increasingly used as a means for ‘‘probing’’ neural models to assess their knowledge of arbitrary linguistic phenomena (Dasgupta et al., 2018; Ettinger et al., 2018; Poliak et al., 2018b; White et al., 2017; Poliak et al., 2018a; McCoy et al., 2019).”
Grammatical Analysis of Pretrained Sentence Encoders with Acceptability Judgments (2019, background): “Ettinger et al. (2018) Allyson Ettinger, Ahmed Elgohary, Colin Phillips, and Philip Resnik.”
What Happens To BERT Embeddings During Fine-tuning? (2020, background): “behavioral studies with challenge sets (McCoy et al., 2019b; Poliak et al., 2018; Ettinger et al., 2018; Kim et al., 2018) have shown limited ability to generalize to out-of-domain data and across syntactic perturbations.”
Interpretability and Analysis in Neural NLP (2020, methodology): “…inference tasks (Poliak et al., 2018a,b; White et al., 2017; Kim et al., 2019; McCoy et al., 2019; Ross and Pavlick, 2019), psycholinguistic methods (Ettinger et al., 2018; Chrupała and Alishahi, 2019), layerwise analyses (Peters et al., 2018; Tenney et al., 2019a), among other methods (Hewitt…”
Learning to Recognize Dialect Features (2020, methodology): “Minimal pairs have also been used to design controlled experiments and probe neural models’ ability to capture various linguistic phenomena (Gulordava et al., 2018; Ettinger et al., 2018; Futrell et al., 2019; Gardner et al., 2020; Schuster et al., 2020).”
Learning to Learn Words from Visual Scenes (2019, background): “Compositions have also been studied in natural language processing [9,12].”
Refining Targeted Syntactic Evaluation of Language Models (2021, background): “…Syntactic Evaluation (TSE), allows for the fine-grained evaluation of models on specific, often rare, syntactic phenomena (Marvin and Linzen, 2018; Ettinger et al., 2018; Warstadt et al., 2020), but (when evaluating S/V number agreement) relies on researchers hand-specifying a small set of verb…”
The EOS Decision and Length Extrapolation (2020, background): “This linear-separability suggests we can train a linear classiﬁer on top of model hidden states, a probe, to predict where sequences can end (Conneau et al., 2018; Ettinger et al., 2018).”
RNNs Implicitly Implement Tensor Product Representations (2018, unknown): “Another approach is based on probing tasks, which assess what information can be easily decoded from a vector representation (Shi et al. 2016; Belinkov et al. 2017; Kádár et al. 2017; Ettinger et al. 2018; compare work in cognitive neuroscience, e.g. Norman et al. 2006).”
Studying the Inductive Biases of RNNs with Synthetic Variations of Natural Languages (2019, background): “What’s more, some sentences are truly ambiguous when there are no case markers or word order cues; this happens for example when both arguments are animate, as in the lawyer saw the doctor (Gibson et al., 2013; Ettinger et al., 2018).”