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.

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

We present minicons, an open source library that provides a standard API for researchers interested in conducting behavioral and representational analyses of transformer-based language models (LMs). Specifically, minicons enables researchers to apply analysis methods at two levels: (1) at the prediction level -- by providing functions to efficiently extract word/sentence level probabilities; and (2) at the representational level -- by also facilitating efficient extraction of word/phrase level vectors from one or more layers. In this paper, we describe the library and apply it to two motivating case studies: One focusing on the learning dynamics of the BERT architecture on relative grammatical judgments, and the other on benchmarking 23 different LMs on zero-shot abductive reasoning. minicons is available at https://github.com/kanishkamisra/minicons

minicons: Enabling Flexible Behavioral and Representational Analyses of Transformer Language Models

Linguistically informed analyses of language models (LMs) contribute to the understanding and improvement of such models. Here, we introduce the corpus of Chinese linguistic minimal pairs (CLiMP) to investigate what knowledge Chinese LMs acquire. CLiMP consists of sets of 1000 minimal pairs (MPs) for 16 syntactic contrasts in Chinese, covering 9 major Chinese linguistic phenomena. The MPs are semi-automatically generated, and human agreement with the labels in CLiMP is 95.8%. We evaluate 11 different LMs on CLiMP, covering n-grams, LSTMs, and Chinese BERT. We find that classifier–noun agreement and verb complement selection are the phenomena that models generally perform best at. However, models struggle the most with the ba construction, binding, and filler-gap dependencies. Overall, Chinese BERT achieves an 81.8% average accuracy, while the performances of LSTMs and 5-grams are only moderately above chance level.

/pdf/climp-a-benchmark-for-chinese-language-model-evaluation-1tbyk61006.pdf

CLiMP: A Benchmark for Chinese Language Model Evaluation

Corpus analysis can be expanded and scaled up by incorporating computational methods from natural language processing. This Element shows how text classification and text similarity models can extend our ability to undertake corpus linguistics across very large corpora. These computational methods are becoming increasingly important as corpora grow too large for more traditional types of linguistic analysis. We draw on five case studies to show how and why to use computational methods, ranging from usage-based grammar to authorship analysis to using social media for corpus-based sociolinguistics. Each section is accompanied by an interactive code notebook that shows how to implement the analysis in Python. A stand-alone Python package is also available to help readers use these methods with their own data. Because large-scale analysis introduces new ethical problems, this Element pairs each new methodology with a discussion of potential ethical implications.

Natural Language Processing for Corpus Linguistics

Relations between words are governed by hierarchical structure rather than linear ordering. Sequence-to-sequence (seq2seq) models, despite their success in downstream NLP applications, often fail to generalize in a hierarchy-sensitive manner when performing syntactic transformations—for example, transforming declarative sentences into questions. However, syntactic evaluations of seq2seq models have only observed models that were not pre-trained on natural language data before being trained to perform syntactic transformations, in spite of the fact that pre-training has been found to induce hierarchical linguistic generalizations in language models; in other words, the syntactic capabilities of seq2seq models may have been greatly understated. We address this gap using the pre-trained seq2seq models T5 and BART, as well as their multilingual variants mT5 and mBART. We evaluate whether they generalize hierarchically on two transformations in two languages: question formation and passivization in English and German. We find that pre-trained seq2seq models generalize hierarchically when performing syntactic transformations, whereas models trained from scratch on syntactic transformations do not. This result presents evidence for the learnability of hierarchical syntactic information from non-annotated natural language text while also demonstrating that seq2seq models are capable of syntactic generalization, though only after exposure to much more language data than human learners receive.

Coloring the Blank Slate: Pre-training Imparts a Hierarchical Inductive Bias to Sequence-to-sequence Models

A range of studies have concluded that neural word prediction models can distinguish grammatical from ungrammatical sentences with high accuracy. However, these studies are based primarily on monolingual evidence from English. To investigate how these models' ability to learn syntax varies by language, we introduce CLAMS (Cross-Linguistic Assessment of Models on Syntax), a syntactic evaluation suite for monolingual and multilingual models. CLAMS includes subject-verb agreement challenge sets for English, French, German, Hebrew and Russian, generated from grammars we develop. We use CLAMS to evaluate LSTM language models as well as monolingual and multilingual BERT. Across languages, monolingual LSTMs achieved high accuracy on dependencies without attractors, and generally poor accuracy on agreement across object relative clauses. On other constructions, agreement accuracy was generally higher in languages with richer morphology. Multilingual models generally underperformed monolingual models. Multilingual BERT showed high syntactic accuracy on English, but noticeable deficiencies in other languages.

/pdf/cross-linguistic-syntactic-evaluation-of-word-prediction-3n5af3gow2.pdf

Panayiota Petrou-Zeniou

Papers

Cross-Linguistic Syntactic Evaluation of Word Prediction Models.

Cross-Linguistic Syntactic Evaluation of Word Prediction Models.