Search for: All records

Because meaning can often be inferred from lexical semantics alone, word order is often a redundant cue in natural language. For example, the words chopped, chef, and onion are more likely used to convey “The chef chopped the onion,” not “The onion chopped the chef.” Recent work has shown large language models to be surprisingly word order invariant, but crucially has largely considered natural prototypical inputs, where compositional meaning mostly matches lexical expectations. To overcome this confound, we probe grammatical role representation in English BERT and GPT-2, on instances where lexical expectations are not sufficient, and word order knowledge is necessary for correct classification. Such non-prototypical instances are naturally occurring English sentences with inanimate subjects or animate objects, or sentences where we systematically swap the arguments to make sentences like “The onion chopped the chef”. We find that, while early layer embeddings are largely lexical, word order is in fact crucial in defining the later-layer representations of words in semantically non-prototypical positions. Our experiments isolate the effect of word order on the contextualization process, and highlight how models use context in the uncommon, but critical, instances where it matters. 

We investigate how to use pretrained static word embeddings to deliver improved estimates of bilexical co-occurrence probabilities: conditional probabilities of one word given a single other word in a specific relationship. Such probabilities play important roles in psycholinguistics, corpus linguistics, and usage-based cognitive modeling of language more generally. We propose a log-bilinear model taking pretrained vector representations of the two words as input, enabling generalization based on the distributional information contained in both vectors. We show that this model outperforms baselines in estimating probabilities of adjectives given nouns that they attributively modify, and probabilities of nominal direct objects given their head verbs, given limited training data in Arabic, English, Korean, and Spanish. 

Languages often express grammatical information through inflectional morphology, in which grammatical features are grouped into strings of morphemes. In this work, we propose that cross-linguistic generalizations about morphological fusion, in which multiple features are expressed through one morpheme, can be explained in part by optimization of processing efficiency, as formalized using the memory--surprisal tradeoff of Hahn et al. (2021). We show in a toy setting that fusion of highly informative neighboring morphemes can lead to greater processing efficiency under our processing model. Next, based on paradigm and frequency data from four languages, we consider both total fusion and gradable fusion using empirical measures developed by Rathi et al. (2021), and find that the degree of fusion is predicted by closeness of optimal morpheme ordering as determined by optimization of processing efficiency. Finally, we show that optimization of processing efficiency can successfully predict typological patterns involving suppletion. 

We investigate how Multilingual BERT (mBERT) encodes grammar by examining how the high-order grammatical feature of morphosyntactic alignment (how different languages define what counts as a “subject”) is manifested across the embedding spaces of different languages. To understand if and how morphosyntactic alignment affects contextual embedding spaces, we train classifiers to recover the subjecthood of mBERT embeddings in transitive sentences (which do not contain overt information about morphosyntactic alignment) and then evaluate them zero-shot on intransitive sentences (where subjecthood classification depends on alignment), within and across languages. We find that the resulting classifier distributions reflect the morphosyntactic alignment of their training languages. Our results demonstrate that mBERT representations are influenced by high-level grammatical features that are not manifested in any one input sentence, and that this is robust across languages. Further examining the characteristics that our classifiers rely on, we find that features such as passive voice, animacy and case strongly correlate with classification decisions, suggesting that mBERT does not encode subjecthood purely syntactically, but that subjecthood embedding is continuous and dependent on semantic and discourse factors, as is proposed in much of the functional linguistics literature. Together, these results provide insight into how grammatical features manifest in contextual embedding spaces, at a level of abstraction not covered by previous work. 

Abstract It is now a common practice to compare models of human language processing by comparing how well they predict behavioral and neural measures of processing difficulty, such as reading times, on corpora of rich naturalistic linguistic materials. However, many of these corpora, which are based on naturally-occurring text, do not contain many of the low-frequency syntactic constructions that are often required to distinguish between processing theories. Here we describe a new corpus consisting of English texts edited to contain many low-frequency syntactic constructions while still sounding fluent to native speakers. The corpus is annotated with hand-corrected Penn Treebank-style parse trees and includes self-paced reading time data and aligned audio recordings. We give an overview of the content of the corpus, review recent work using the corpus, and release the data. 

Linguistic typology generally divides synthetic languages into groups based on their morphological fusion. However, this measure has long been thought to be best considered a matter of degree. We present an information-theoretic measure, called informational fusion, to quantify the degree of fusion of a given set of morphological features in a surface form, which naturally provides such a graded scale. Informational fusion is able to encapsulate not only concatenative, but also nonconcatenative morphological systems (e.g. Arabic), abstracting away from any notions of morpheme segmentation. We then show, on a sample of twenty-one languages, that our measure recapitulates the usual linguistic classifications for concatenative systems, and provides new measures for nonconcatenative ones. We also evaluate the long-standing hypotheses that more frequent forms are more fusional, and that paradigm size anticorrelates with degree of fusion. We do not find evidence for the idea that languages have characteristic levels of fusion; rather, the degree of fusion varies across part-of-speech within languages. 

Abstract We introduce a theoretical framework for understanding and predicting the complexity of sequence classification tasks, using a novel extension of the theory of Boolean function sensitivity. The sensitivity of a function, given a distribution over input sequences, quantifies the number of disjoint subsets of the input sequence that can each be individually changed to change the output. We argue that standard sequence classification methods are biased towards learning low-sensitivity functions, so that tasks requiring high sensitivity are more difficult. To that end, we show analytically that simple lexical classifiers can only express functions of bounded sensitivity, and we show empirically that low-sensitivity functions are easier to learn for LSTMs. We then estimate sensitivity on 15 NLP tasks, finding that sensitivity is higher on challenging tasks collected in GLUE than on simple text classification tasks, and that sensitivity predicts the performance both of simple lexical classifiers and of vanilla BiLSTMs without pretrained contextualized embeddings. Within a task, sensitivity predicts which inputs are hard for such simple models. Our results suggest that the success of massively pretrained contextual representations stems in part because they provide representations from which information can be extracted by low-sensitivity decoders.