Probing What Different NLP Tasks Teach Machines about Function Word Comprehension

By Najoung Kim, Roma Patel, Adam Poliak, Alex Wang, Patrick Xia, R. Thomas McCoy, Ian Tenney, Alexis Ross, Tal Linzen, Benjamin Van Durme and 2 more. 2019

115 citations recorded by the source corpus.

We introduce a set of nine challenge tasks that test for the understanding of function words. These tasks are created by structurally mutating sentences from existing datasets to target the comprehension of specific types of function words (e.g., prepositions, wh-words). Using these probing tasks, we explore the effects of various pretraining objectives for sentence encoders (e.g., language modeling, CCG supertagging and natural language inference (NLI)) on the learned representations. Our results show that pretraining on CCG—our most syntactic objective—performs the best on average across our probing tasks, suggesting that syntactic knowledge helps function word comprehension. Language modeling also shows strong performance, supporting its widespread use for pretraining state-of-the-art NLP models. Overall, no pretraining objective dominates across the board, and our function word probing tasks highlight several intuitive differences between pretraining objectives, e.g., that NLI helps the comprehension of negation.

Subfields: Artificial Intelligence, 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 Natural Language Processing

Citations in the corpus

Polyjuice: Generating Counterfactuals for Explaining, Evaluating, and Improving Models (2021, methodology): “m-polyjuice improves accuracy on contrast and challenge sets (Kim et al., 2019; Naik et al., 2018; Glockner et al., 2018; Wang et al., 2019); it exhibits comparable (or better) gains than m-CAD (manual counterfactuals) with less implementation and annotation effort.”
Beyond Accuracy: Behavioral Testing of NLP Models with CheckList (2020, methodology): “NER) is trained on intermediate representations of the encoder (Tenney et al., 2019; Kim et al., 2019).”
Designing and Interpreting Probes with Control Tasks (2019, methodology): “Applying dropout, the most popular probing regularization method (Adi et al., 2017; Belinkov and Glass, 2019; Şahin et al., 2019; Kim et al., 2019; Elloumi et al., 2018; Belinkov and Glass, 2017; Belinkov et al., 2018) does not consistently lead to high-accuracy, high-selectivity MLP probes across a”
Negated and Misprimed Probes for Pretrained Language Models: Birds Can Talk, But Cannot Fly (2019, background): “Kim et al. (2019) investigate understanding of negation particles when PLMs are finetuned.”
jiant: A Software Toolkit for Research on General-Purpose Text Understanding Models (2020, background): “…including probing analyses (Ten-ney et al., 2019b,a; Warstadt et al., 2019; Lin et al., 2019; Hewitt and Manning, 2019; Jawahar et al., 2019), transfer learning experiments (Wang et al., 2019a; Phang et al., 2018), and dataset and benchmark construction (Wang et al., 2019b, 2018; Kim et al., 2019).”
BERTs of a feather do not generalize together: Large variability in generalization across models with similar test set performance (2019, background): “Many recent papers have sought a deeper understanding of BERT, whether to assess its encoding of sentence structure (Lin et al., 2019; Hewitt and Manning, 2019; Chrupała and Alishahi, 2019; Jawahar et al., 2019; Tenney et al., 2019b); its representational structure more generally (Abnar et al.,…”
Intermediate-Task Transfer Learning with Pretrained Language Models: When and Why Does It Work? (2020, methodology): “We use well-established datasets for our probing tasks, including the edge-probing suite from Ten-ney et al. (2019b), function word oriented tasks from Kim et al. (2019), and sentence-level probing datasets (SentEval; Conneau et al., 2018).”
Evaluating Commonsense in Pre-trained Language Models (2019, background): “There has been work investigating the nature of syntactic (Liu et al. 2019a), semantic (Liu et al. 2019a) and word sense (Kim et al. 2019) knowledge contained in such contextualized representations, in particular BERT (Devlin et al. 2019), showing ∗Work done while at Westlake University Copyright…”
TACRED Revisited: A Thorough Evaluation of the TACRED Relation Extraction Task (2020, unknown): “…meth-ods include occlusion or gradient-based methods, measuring the relevance of input features to the output (Zintgraf et al., 2017; Harbecke et al., 2018), and probing tasks (Conneau et al., 2018; Kim et al., 2019) that probe the presence of speciﬁc features e.g. in intermediate layers.”
When a sentence does not introduce a discourse entity, Transformer-based models still sometimes refer to it (2022, methodology): “Kim et al. (2019) used an acceptability judgment task to determine whether contextual LMs correctly distinguish between deﬁnite and indeﬁnite noun phrases.”
Sometimes We Want Ungrammatical Translations (2021, background): “For specific NLP tasks, probe tasks are engineered to measure specific kinds of linguistic knowledge encoded in the systems (Conneau et al., 2018; Sheng et al., 2019; Kim et al., 2019; Jeretic et al., 2020; Parthasarathi et al., 2020; Ribeiro et al., 2020).”
DirectProbe: Studying Representations without Classifiers (2021, background): “Commonly measured properties include generalization error (Kim et al., 2019), minimum description length of labels (Voita and Titov, 2020) and complexity (Whitney et al., 2020).”
CONDAQA: A Contrastive Reading Comprehension Dataset for Reasoning about Negation (2022, background): “, 2021), NLI data (Kim et al., 2019), commonsense knowledge (Jiang et al.”
Probing the Probing Paradigm: Does Probing Accuracy Entail Task Relevance? (2020, background): “Kim et al. (2019) study what different NLP tasks teach models about function word comprehension.”
Multi-task learning for natural language processing in the 2020s: where are we going? (2020, background): “A host of other researchers have gone on to argue that task re-latedness plays a key role in determining how knowledge information is shared (Mou et al., 2016; Ben-David and Borbely, 2008) and when an auxiliary task will help and when it will hurt (Standley et al., 2019; Kim et al., 2019).”
On the Systematicity of Probing Contextualized Word Representations: The Case of Hypernymy in BERT (2020, background): “Much recent work focuses on the grammatical and syntactic capabilities of BERT (Hewitt and Manning, 2019; Liu et al., 2019; Swayamdipta et al., 2019; Goldberg, 2019; Wolf, 2019; Coenen et al., 2019; Tenney et al., 2019; Warstadt et al., 2019; Kim et al., 2019).”
Negated LAMA: Birds cannot fly (2019, methodology): “Kim et al. (2019) investigate the understanding of function words – among them negation particles – using an entailment-and classiﬁcation-based approach.”