Neural Network Acceptability Judgments Warstadt, Alex; Singh, Amanpreet; Bowman, Samuel R.
Transactions of the Association for Computational Linguistics,
11/2019, Letnik:
7
Journal Article
Recenzirano
Odprti dostop
This paper investigates the ability of artificial neural networks to judge the grammatical acceptability of a sentence, with the goal of testing their linguistic competence. We introduce the Corpus ...of Linguistic Acceptability (CoLA), a set of 10,657 English sentences labeled as grammatical or ungrammatical from published linguistics literature. As baselines, we train several recurrent neural network models on acceptability classification, and find that our models outperform unsupervised models by Lau et al. (2016) on CoLA. Error-analysis on specific grammatical phenomena reveals that both Lau et al.’s models and ours learn systematic generalizations like subject-verb-object order. However, all models we test perform far below human level on a wide range of grammatical constructions.
We introduce The Benchmark of Linguistic Minimal Pairs (BLiMP),
a challenge set for evaluating the linguistic knowledge of language models (LMs) on major grammatical phenomena in English. BLiMP ...consists of 67 individual datasets, each containing 1,000 minimal pairs—that is, pairs of minimally different sentences that contrast in grammatical acceptability and isolate specific phenomenon in syntax, morphology, or semantics. We generate the data according to linguist-crafted grammar templates, and human aggregate agreement with the labels is 96.4%. We evaluate
-gram, LSTM, and Transformer (GPT-2 and Transformer-XL) LMs by observing whether they assign a higher probability to the acceptable sentence in each minimal pair. We find that state-of-the-art models identify morphological contrasts related to agreement reliably, but they struggle with some subtle semantic and syntactic phenomena, such as negative polarity items and extraction islands.
Data-driven learning uncontroversially plays a role in human language acquisition---how large a role is a matter of much debate. The success of artificial neural networks in NLP in recent years calls ...for a re-evaluation of our understanding of the possibilities for learning grammar from data alone. This dissertation argues the case for using artificial neural networks to test hypotheses about human language acquisition and presents progress towards this goal from multiple directions. Compared to experiments on human subjects, experiments on artificial learners based on neural networks have massive advantages in terms of ethics, expense, and expanded possibilities for experimental design. I provide a general recipe for building more convincing model learners and designing ablation experiments using them (Chapter 1). Subsequently, I introduce benchmarks including The Corpus of Linguistic Acceptability (CoLA; Chapter 2) and The Benchmark of Linguistic Minimal Pairs for English (BLiMP; Chapter 3) that use acceptability judgments to probe grammatical knowledge in artificial neural networks. Although off-the-shelf neural language models popular in natural language processing today achieve human-level performance on these benchmarks, they are trained on orders of magnitude more linguistic input than children are exposed to, making them unsuitable for studying human language acquisition. Thus, I train language models from scratch in more human-like settings using limited data, and track the acquisition of language-specific inductive bias and grammatical features as a function of the volume of input (Chapters 4 and 5). Results show that there is a remarkably rich signal for grammar learning in raw text data, but current models require considerably more data than a child to learn from it. Chapter 6 is a synthesis of these approaches applied to a long-standing debate in language acquisition regarding the learnability of structure dependence in subject auxiliary inversion. Through a controlled experimental manipulation of the learning environment of neural language models, I find evidence that in the absence of an innate hierarchical bias, direct evidence against a linear rule, while helpful, is not necessary in order for data-driven learners to learn acceptability judgments consistent with a structural generalization. These results not only highlight the importance of considering indirect evidence in learnability debates, but provide a proof-of-concept for the use artificial learners in evaluating previously untestable hypotheses about human language acquisition.
In contrast to children, language models (LMs) exhibit considerably inferior data efficiency when acquiring language. In this submission to the BabyLM Challenge (Warstadt et al., 2023), we test the ...hypothesis that this data efficiency gap is partly caused by a lack of multimodal input and grounding in the learning environment of typical language models. Although previous work looking into this question found that multimodal training can even harm language-only performance, we speculate that these findings can be attributed to catastrophic forgetting of complex language due to fine-tuning on captions data. To test our hypothesis, we perform an ablation study on FLAVA (Singh et al., 2022), a multimodal vision-and-language model, independently varying the volume of text and vision input to quantify how much text data (if any) can be offset by vision at different data scales. We aim to limit catastrophic forgetting through a multitask pretraining regime that includes unimodal text-only tasks and data sampled from WiT, the relatively diverse Wikipedia-based dataset (Srinivasan et al., 2021). Our results are largely negative: Multimodal pretraining does not harm our models' language performance but does not consistently help either. That said, our conclusions are limited by our having been able to conduct only a small number of runs. While we must leave open the possibility that multimodal input explains some of the gap in data efficiency between LMs and humans, positive evidence for this hypothesis will require better architectures and techniques for multimodal training.
Rapid progress in machine learning for natural language processing has the potential to transform debates about how humans learn language. However, the learning environments and biases of current ...artificial learners and humans diverge in ways that weaken the impact of the evidence obtained from learning simulations. For example, today's most effective neural language models are trained on roughly one thousand times the amount of linguistic data available to a typical child. To increase the relevance of learnability results from computational models, we need to train model learners without significant advantages over humans. If an appropriate model successfully acquires some target linguistic knowledge, it can provide a proof of concept that the target is learnable in a hypothesized human learning scenario. Plausible model learners will enable us to carry out experimental manipulations to make causal inferences about variables in the learning environment, and to rigorously test poverty-of-the-stimulus-style claims arguing for innate linguistic knowledge in humans on the basis of speculations about learnability. Comparable experiments will never be possible with human subjects due to practical and ethical considerations, making model learners an indispensable resource. So far, attempts to deprive current models of unfair advantages obtain sub-human results for key grammatical behaviors such as acceptability judgments. But before we can justifiably conclude that language learning requires more prior domain-specific knowledge than current models possess, we must first explore non-linguistic inputs in the form of multimodal stimuli and multi-agent interaction as ways to make our learners more efficient at learning from limited linguistic input.
Language models are often trained on text alone, without additional grounding. There is debate as to how much of natural language semantics can be inferred from such a procedure. We prove that ...entailment judgments between sentences can be extracted from an ideal language model that has perfectly learned its target distribution, assuming the training sentences are generated by Gricean agents, i.e., agents who follow fundamental principles of communication from the linguistic theory of pragmatics. We also show entailment judgments can be decoded from the predictions of a language model trained on such Gricean data. Our results reveal a pathway for understanding the semantic information encoded in unlabeled linguistic data and a potential framework for extracting semantics from language models.
Humans appear to have a critical period (CP) for language acquisition: Second language (L2) acquisition becomes harder after early childhood, and ceasing exposure to a first language (L1) after this ...period (but not before) typically does not lead to substantial loss of L1 proficiency. It is unknown whether these CP effects result from innately determined brain maturation or as a stabilization of neural connections naturally induced by experience. In this study, we use language models (LMs) to test the extent to which these phenomena are peculiar to humans, or shared by a broader class of language learners. We vary the age of exposure by training LMs on language pairs in various experimental conditions, and find that LMs, which lack any direct analog to innate maturational stages, do not show CP effects when trained sequentially on L1 and L2. Our results contradict the claim that CP effects are an inevitable result of learning in statistical learners, and they are consistent with an innate mechanism for CP effects. We show that we can reverse-engineer the CP by introducing a regularizer partway through training to simulate a maturational decrease in plasticity. All in all, our results suggest that L1 learning on its own may not be enough to induce a CP, and additional engineering is necessary to make language models more cognitively plausible.
We evaluate whether BERT, a widely used neural network for sentence processing, acquires an inductive bias towards forming structural generalizations through pretraining on raw data. We conduct four ...experiments testing its preference for structural vs. linear generalizations in different structure-dependent phenomena. We find that BERT makes a structural generalization in 3 out of 4 empirical domains---subject-auxiliary inversion, reflexive binding, and verb tense detection in embedded clauses---but makes a linear generalization when tested on NPI licensing. We argue that these results are the strongest evidence so far from artificial learners supporting the proposition that a structural bias can be acquired from raw data. If this conclusion is correct, it is tentative evidence that some linguistic universals can be acquired by learners without innate biases. However, the precise implications for human language acquisition are unclear, as humans learn language from significantly less data than BERT.
Recent work on evaluating grammatical knowledge in pretrained sentence encoders gives a fine-grained view of a small number of phenomena. We introduce a new analysis dataset that also has broad ...coverage of linguistic phenomena. We annotate the development set of the Corpus of Linguistic Acceptability (CoLA; Warstadt et al., 2018) for the presence of 13 classes of syntactic phenomena including various forms of argument alternations, movement, and modification. We use this analysis set to investigate the grammatical knowledge of three pretrained encoders: BERT (Devlin et al., 2018), GPT (Radford et al., 2018), and the BiLSTM baseline from Warstadt et al. We find that these models have a strong command of complex or non-canonical argument structures like ditransitives (Sue gave Dan a book) and passives (The book was read). Sentences with long distance dependencies like questions (What do you think I ate?) challenge all models, but for these, BERT and GPT have a distinct advantage over the baseline. We conclude that recent sentence encoders, despite showing near-human performance on acceptability classification overall, still fail to make fine-grained grammaticality distinctions for many complex syntactic structures.