Discourse structure interacts with reference but not syntax in neural language models

Abstract

Language models (LMs) trained on large quantities of text have been claimed to acquire abstract linguistic representations. Our work tests the robustness of these abstractions by focusing on the ability of LMs to learn interactions between different linguistic representations. In particular, we utilized stimuli from psycholinguistic studies showing that humans can condition reference (i.e. coreference resolution) and syntactic processing on the same discourse structure (implicit causality). We compared both transformer and long short-term memory LMs to find that, contrary to humans, implicit causality only influences LM behavior for reference, not syntax, despite model representations that encode the necessary discourse information. Our results further suggest that LM behavior can contradict not only learned representations of discourse but also syntactic agreement, pointing to shortcomings of standard language modeling.

Uncovering Constraint-Based Behavior in Neural Models via Targeted Fine-Tuning

“…Prior work has noted competing generalizations influencing model behavior via the distinction of non-linguistic vs. linguistic biases (e.g., Mc-Coy et al, 2019;Davis and van Schijndel, 2020a;Warstadt et al, 2020b). The findings in Warstadt et al (2020b), that linguistic knowledge is represented within a model much earlier than attestation in model behavior, bears resemblance to our claims.…”

“…Given that all the investigated stimuli were disambiguated by gender, we categorized our results by the antecedent of the pronoun and the IC verb bias. We first turn to English and Chinese, which showed an IC bias in line with existing work on IC bias in autoregressive English models (e.g., Upadhye et al, 2020;Davis and van Schijndel, 2020a). We then detail the results for Spanish and Italian, where only very limited, if any, IC bias was observed.…”

“…Current accounts of IC ground the phenomenon within the linguistic signal without the need for additional pragmatic inferences by comprehenders (e.g., Ro-hde et al, 2011;Hartshorne et al, 2013). Recent investigations of IC in neural language models confirms that the IC bias of English is learnable, at least to some degree, from text data alone (Davis and van Schijndel, 2020a;Upadhye et al, 2020). The ability of models trained on other languages to acquire an IC bias, however, has not been explored.…”

“…Thus, the apparent failure of the Spanish and Italian models to pattern like English and Chinese is not evidence on its own of a model's inability to acquire the requisite linguistic 1 These model types were chosen for ease of access to existing models. Pretrained, large auto-regressive models are largely restricted to English, and prior work suggests that LSTMs are limited in their ability to acquire an IC bias in English (Davis and van Schijndel, 2020a).…”

John praised Mary because _he_? Implicit Causality Bias and Its Interaction with Explicit Cues in LMs

“…While they did not find strong evidence for a correlation to human-based results in this respect, they did observe that in the context of connective because PLMs assigned lower probability to subjectreferring pronouns for an object-biasing verb as compared to a subject-biasing verb. Davis and van Schijndel (2020) observed that GPT2-XL (Radford et al, 2019) encodes some level of IC bias in its representations (measured in terms of similarity between the representation of the pronoun and its two potential referents) and its decision on how to resolve a referent at prediction time is weakly influenced by that. They took the analysis one step further and looked at whether GPT2-XL uses IC information to resolve relative clause attachment, which in humans is conditioned by IC bias-no evidence was found to suggest that that was the case.…”