Recent studies have shown that as Transformer-based language models become larger and are trained on very large amounts of data, the fit of their surprisal estimates to naturalistic human reading times degrades. The current work presents a series of analyses showing that word frequency is a key explanatory factor underlying these two trends. First, residual errors from four language model families on four corpora show that the inverse correlation between model size and fit to reading times is the strongest on the subset of least frequent words, which is driven by excessively accurate predictions of larger model variants. Additionally, training dynamics reveal that during later training steps, all model variants learn to predict rare words and that larger model variants do so more accurately, which explains the detrimental effect of both training data amount and model size on fit to reading times. Finally, a feature attribution analysis demonstrates that larger model variants are able to accurately predict rare words based on both an effectively longer context window size as well as stronger local associations compared to smaller model variants. Taken together, these results indicate that Transformer-based language models’ surprisal estimates diverge from human-like expectations due to the superhumanly complex associations they learn for predicting rare words.
Grammar induction, the task of learning a set of syntactic rules from minimally annotated training data, provides a means of exploring the longstanding question of whether humans rely on innate knowledge to acquire language. Of the various formalisms available for grammar induction, categorial grammars provide an appealing option due to their transparent interface between syntax and semantics. However, to obtain competitive results, previous categorial grammar inducers have relied on shortcuts such as part-of-speech annotations or an ad hoc bias term in the objective function to ensure desirable branching behavior. We present a categorial grammar inducer that eliminates both shortcuts: it learns from raw data, and does not rely on a biased objective function. This improvement is achieved through a novel stochastic process used to select the set of available syntactic categories. On a corpus of English child-directed speech, the model attains a recall-homogeneity of 0.48, a large improvement over previous categorial grammar inducers.
This work presents a linguistic analysis into why larger Transformer-based pre-trained language models with more parameters and lower perplexity nonetheless yield surprisal estimates that are less predictive of human reading times. First, regression analyses show a strictly monotonic, positive log-linear relationship between perplexity and fit to reading times for the more recently released five GPT-Neo variants and eight OPT variants on two separate datasets, replicating earlier results limited to just GPT-2 (Oh et al., 2022). Subsequently, analysis of residual errors reveals a systematic deviation of the larger variants, such as underpredicting reading times of named entities and making compensatory overpredictions for reading times of function words such as modals and conjunctions. These results suggest that the propensity of larger Transformer-based models to ‘memorize’ sequences during training makes their surprisal estimates diverge from humanlike expectations, which warrants caution in using pre-trained language models to study human language processing.
Creating conversational systems for niche domains is a challenging task, further exacerbated by a lack of quality datasets. We explore the construction of safer conversational systems for guiding patients in preparing for colonoscopies. This has required a data generation pipeline to generate a minimum viable dataset to bootstrap a semantic parser, augmented by automatic paraphrasing. Our study suggests large language models (e.g., GPT-3.5 and GPT-4) are a viable alternative to crowd sourced paraphrasing, but conversational systems that rely upon language models’ ability to do temporal reasoning struggle to provide accurate responses. A neural-symbolic system that performs temporal reasoning on an intermediate representation of user queries shows promising results compared to an end-to-end dialogue system, improving the number of correct responses while vastly reducing the number of incorrect or misleading ones.
Grammar induction, the task of learning a set of grammatical rules from raw or minimally labeled text data, can provide clues about what kinds of syntactic structures are learnable without prior knowledge. Recent work (e.g., Kim et al., 2019; Zhu et al., 2020; Jin et al., 2021a) has achieved advances in unsupervised induction of probabilistic context-free grammars (PCFGs). However, categorial grammar induction has received less recent attention, despite allowing inducers to support a larger set of syntactic categories—due to restrictions on how categories can combine—and providing a transparent interface with compositional semantics, opening up possibilities for models that jointly learn form and meaning. Motivated by this, we propose a new model for inducing a basic (Ajdukiewicz, 1935; Bar-Hillel, 1953) categorial grammar. In contrast to earlier categorial grammar induction systems (e.g., Bisk and Hockenmaier, 2012), our model learns from raw data without any part-of-speech information. Experiments on child-directed speech show that our model attains a recall-homogeneity of 0.33 on average, which dramatically increases to 0.59 when a bias toward forward function application is added to the model.
Recent psycholinguistic studies have drawn conflicting conclusions about the relationship between the quality of a language model and the ability of its surprisal estimates to predict human reading times, which has been speculated to be due to the large gap in both the amount of training data and model capacity across studies. The current work aims to consolidate these findings by evaluating surprisal estimates from Transformer-based language model variants that vary systematically in the amount of training data and model capacity on their ability to predict human reading times. The results show that surprisal estimates from most variants with contemporary model capacities provide the best fit after seeing about two billion training tokens, after which they begin to diverge from humanlike expectations. Additionally, newly-trained smaller model variants reveal a ‘tipping point’ at convergence, after which the decrease in language model perplexity begins to result in poorer fits to human reading times. These results suggest that the massive amount of training data is mainly responsible for the poorer fit achieved by surprisal from larger pre-trained language models, and that a certain degree of model capacity is necessary for Transformer-based language models to capture humanlike expectations.
While there is much recent interest in studying why Transformer-based large language models make predictions the way they do, the complex computations performed within each layer have made their behavior somewhat opaque. To mitigate this opacity, this work presents a linear decomposition of final hidden states from autoregressive language models based on each initial input token, which is exact for virtually all contemporary Transformer architectures. This decomposition allows the definition of probability distributions that ablate the contribution of specific input tokens, which can be used to analyze their influence on model probabilities over a sequence of upcoming words with only one forward pass from the model. Using the change in next-word probability as a measure of importance, this work first examines which context words make the biggest contribution to language model predictions. Regression experiments suggest that Transformer-based language models rely primarily on collocational associations, followed by linguistic factors such as syntactic dependencies and coreference relationships in making next-word predictions. Additionally, analyses using these measures to predict syntactic dependencies and coreferent mention spans show that collocational association and repetitions of the same token largely explain the language models’ predictions on these tasks.
Transformer-based large language models are trained to make predictions about the next word by aggregating representations of previous tokens through their self-attention mechanism. In the field of cognitive modeling, such attention patterns have recently been interpreted as embodying the process of cue-based retrieval, in which attention over multiple targets is taken to generate interference and latency during retrieval. Under this framework, this work first defines an entropy-based predictor that quantifies the diffuseness of self-attention, as well as distance-based predictors that capture the incremental change in attention patterns across timesteps. Moreover, following recent studies that question the informativeness of attention weights, we also experiment with alternative methods for incorporating vector norms into attention weights. Regression experiments using predictors calculated from the GPT-2 language model show that these predictors deliver a substantially better fit to held-out self-paced reading and eye-tracking data over a rigorous baseline including GPT-2 surprisal.
While the use of character models has been popular in NLP applications, it has not been explored much in the context of psycholinguistic modeling. This paper presents a character model that can be applied to a structural parser-based processing model to calculate word generation probabilities. Experimental results show that surprisal estimates from a structural processing model using this character model deliver substantially better fits to self-paced reading, eye-tracking, and fMRI data than those from large-scale language models trained on much more data. This may suggest that the proposed processing model provides a more humanlike account of sentence processing, which assumes a larger role of morphology, phonotactics, and orthographic complexity than was previously thought.
This article describes a simple PCFG induction model with a fixed category domain that predicts a large majority of attested constituent boundaries, and predicts labels consistent with nearly half of attested constituent labels on a standard evaluation data set of child-directed speech. The article then explores the idea that the difference between simple grammars exhibited by child learners and fully recursive grammars exhibited by adult learners may be an effect of increasing working memory capacity, where the shallow grammars are constrained images of the recursive grammars. An implementation of these memory bounds as limits on center embedding in a depth-specific transform of a recursive grammar yields a significant improvement over an equivalent but unbounded baseline, suggesting that this arrangement may indeed confer a learning advantage.
Recent evidence supports a role for coreference processing in guiding human expectations about upcoming words during reading, based on covariation between reading times and word surprisal estimated by a coreference-aware semantic processing model (Jaffe et al. 2020).The present study reproduces and elaborates on this finding by (1) enabling the parser to process subword information that might better approximate human morphological knowledge, and (2) extending evaluation of coreference effects from self-paced reading to human brain imaging data. Results show that an expectation-based processing effect of coreference is still evident even in the presence of the stronger psycholinguistic baseline provided by the subword model, and that the coreference effect is observed in both self-paced reading and fMRI data, providing evidence of the effect’s robustness.
Unsupervised PCFG induction models, which build syntactic structures from raw text, can be used to evaluate the extent to which syntactic knowledge can be acquired from distributional information alone. However, many state-of-the-art PCFG induction models are word-based, meaning that they cannot directly inspect functional affixes, which may provide crucial information for syntactic acquisition in child learners. This work first introduces a neural PCFG induction model that allows a clean ablation of the influence of subword information in grammar induction. Experiments on child-directed speech demonstrate first that the incorporation of subword information results in more accurate grammars with categories that word-based induction models have difficulty finding, and second that this effect is amplified in morphologically richer languages that rely on functional affixes to express grammatical relations. A subsequent evaluation on multilingual treebanks shows that the model with subword information achieves state-of-the-art results on many languages, further supporting a distributional model of syntactic acquisition.
Expectation-based theories of sentence processing posit that processing difficulty is determined by predictability in context. While predictability quantified via surprisal has gained empirical support, this representation-agnostic measure leaves open the question of how to best approximate the human comprehender’s latent probability model. This work presents an incremental left-corner parser that incorporates information about both propositional content and syntactic categories into a single probability model. This parser can be trained to make parsing decisions conditioning on only one source of information, thus allowing a clean ablation of the relative contribution of propositional content and syntactic category information. Regression analyses show that surprisal estimates calculated from the full parser make a significant contribution to predicting self-paced reading times over those from the parser without syntactic category information, as well as a significant contribution to predicting eye-gaze durations over those from the parser without propositional content information. Taken together, these results suggest a role for propositional content and syntactic category information in incremental sentence processing.
People can extract precise, complex logical meanings from text in documents such as tax forms and game rules, but language processing systems lack adequate training and evaluation resources to do these kinds of tasks reliably. This paper describes a corpus of annotated typed lambda calculus translations for approximately 2,000 sentences in Simple English Wikipedia, which is assumed to constitute a broad-coverage domain for precise, complex descriptions. The corpus described in this paper contains a large number of quantifiers and interesting scoping configurations, and is presented specifically as a resource for quantifier scope disambiguation systems, but also more generally as an object of linguistic study.
Syntactic surprisal has been shown to have an effect on human sentence processing, and can be predicted from prefix probabilities of generative incremental parsers. Recent state-of-the-art incremental generative neural parsers are able to produce accurate parses and surprisal values but have unbounded stack memory, which may be used by the neural parser to maintain explicit in-order representations of all previously parsed words, inconsistent with results of human memory experiments. In contrast, humans seem to have a bounded working memory, demonstrated by inhibited performance on word recall in multi-clause sentences (Bransford and Franks, 1971), and on center-embedded sentences (Miller and Isard,1964). Bounded statistical parsers exist, but are less accurate than neural parsers in predict-ing reading times. This paper describes a neural incremental generative parser that is able to provide accurate surprisal estimates and can be constrained to use a bounded stack. Results show that the accuracy gains of neural parsers can be reliably extended to psycholinguistic modeling without risk of distortion due to un-bounded working memory.
Recent progress in grammar induction has shown that grammar induction is possible without explicit assumptions of language specific knowledge. However, evaluation of induced grammars usually has ignored phrasal labels, an essential part of a grammar. Experiments in this work using a labeled evaluation metric, RH, show that linguistically motivated predictions about grammar sparsity and use of categories can only be revealed through labeled evaluation. Furthermore, depth-bounding as an implementation of human memory constraints in grammar inducers is still effective with labeled evaluation on multilingual transcribed child-directed utterances.
Models of human sentence processing effort tend to focus on costs associated with retrieving structures and discourse referents from memory (memory-based) and/or on costs associated with anticipating upcoming words and structures based on contextual cues (expectation-based) (Levy,2008). Although evidence suggests that expectation and memory may play separable roles in language comprehension (Levy et al., 2013), theories of coreference processing have largely focused on memory: how comprehenders identify likely referents of linguistic expressions. In this study, we hypothesize that coreference tracking also informs human expectations about upcoming words, and we test this hypothesis by evaluating the degree to which incremental surprisal measures generated by a novel coreference-aware semantic parser explain human response times in a naturalistic self-paced reading experiment. Results indicate (1) that coreference information indeed guides human expectations and (2) that coreference effects on memory retrieval may exist independently of coreference effects on expectations. Together, these findings suggest that the language processing system exploits coreference information both to retrieve referents from memory and to anticipate upcoming material.
Recent work in unsupervised parsing has tried to incorporate visual information into learning, but results suggest that these models need linguistic bias to compete against models that only rely on text. This work proposes grammar induction models which use visual information from images for labeled parsing, and achieve state-of-the-art results on grounded grammar induction on several languages. Results indicate that visual information is especially helpful in languages where high frequency words are more broadly distributed. Comparison between models with and without visual information shows that the grounded models are able to use visual information for proposing noun phrases, gathering useful information from images for unknown words, and achieving better performance at prepositional phrase attachment prediction.
Unsupervised PCFG inducers hypothesize sets of compact context-free rules as explanations for sentences. PCFG induction not only provides tools for low-resource languages, but also plays an important role in modeling language acquisition (Bannard et al., 2009; Abend et al. 2017). However, current PCFG induction models, using word tokens as input, are unable to incorporate semantics and morphology into induction, and may encounter issues of sparse vocabulary when facing morphologically rich languages. This paper describes a neural PCFG inducer which employs context embeddings (Peters et al., 2018) in a normalizing flow model (Dinh et al., 2015) to extend PCFG induction to use semantic and morphological information. Linguistically motivated sparsity and categorical distance constraints are imposed on the inducer as regularization. Experiments show that the PCFG induction model with normalizing flow produces grammars with state-of-the-art accuracy on a variety of different languages. Ablation further shows a positive effect of normalizing flow, context embeddings and proposed regularizers.
In unsupervised grammar induction, data likelihood is known to be only weakly correlated with parsing accuracy, especially at convergence after multiple runs. In order to find a better indicator for quality of induced grammars, this paper correlates several linguistically- and psycholinguistically-motivated predictors to parsing accuracy on a large multilingual grammar induction evaluation data set. Results show that variance of average surprisal (VAS) better correlates with parsing accuracy than data likelihood and that using VAS instead of data likelihood for model selection provides a significant accuracy boost. Further evidence shows VAS to be a better candidate than data likelihood for predicting word order typology classification. Analyses show that VAS seems to separate content words from function words in natural language grammars, and to better arrange words with different frequencies into separate classes that are more consistent with linguistic theory.
Researchers in computational psycholinguistics frequently use linear models to study time series data generated by human subjects. However, time series may violate the assumptions of these models through temporal diffusion, where stimulus presentation has a lingering influence on the response as the rest of the experiment unfolds. This paper proposes a new statistical model that borrows from digital signal processing by recasting the predictors and response as convolutionally-related signals, using recent advances in machine learning to fit latent impulse response functions (IRFs) of arbitrary shape. A synthetic experiment shows successful recovery of true latent IRFs, and psycholinguistic experiments reveal plausible, replicable, and fine-grained estimates of latent temporal dynamics, with comparable or improved prediction quality to widely-used alternatives.
There have been several recent attempts to improve the accuracy of grammar induction systems by bounding the recursive complexity of the induction model. Modern depth-bounded grammar inducers have been shown to be more accurate than early unbounded PCFG inducers, but this technique has never been compared against unbounded induction within the same system, in part because most previous depth-bounding models are built around sequence models, the complexity of which grows exponentially with the maximum allowed depth. The present work instead applies depth bounds within a chart-based Bayesian PCFG inducer, where bounding can be switched on and off, and then samples trees with or without bounding. Results show that depth-bounding is indeed significantly effective in limiting the search space of the inducer and thereby increasing accuracy of resulting parsing model, independent of the contribution of modern Bayesian induction techniques. Moreover, parsing results on English, Chinese and German show that this bounded model is able to produce parse trees more accurately than or competitively with state-of-the-art constituency grammar induction models.
There has been recent interest in applying cognitively- or empirically-motivated bounds on recursion depth to limit the search space of grammar induction models (Ponvert et al., 2011; Noji and Johnson, 2016; Shain et al., 2016). This work extends this depth-bounding approach to probabilistic context-free grammar induction (DB-PCFG), which has a smaller parameter space than hierarchical sequence models, and therefore more fully exploits the space reductions of depth-bounding. Results for this model on grammar acquisition from transcribed child-directed speech and newswire text exceed or are competitive with those of other models when evaluated on parse accuracy. Moreover, grammars acquired from this model demonstrate a consistent use of category labels, something which has not been demonstrated by other acquisition models.
This paper presents a new memory-bounded left-corner parsing model for unsupervised raw-text syntax induction, using unsupervised hierarchical hidden Markov models (UHHMM). We deploy this algorithm to shed light on the extent to which human language learners can discover hierarchical syntax through distributional statistics alone, by modeling two widely-accepted features of human language acquisition and sentence processing that have not been simultaneously modeled by any existing grammar induction algorithm: (1) a left-corner parsing strategy and (2) limited working memory capacity. To model realistic input to human language learners, we evaluate our system on a corpus of child-directed speech rather than typical newswire corpora. Results beat or closely match those of three competing systems.
This study demonstrates a weakness in how n-gram and PCFG surprisal are used to predict reading times in eye-tracking data. In particular, the information conveyed by words skipped during saccades is not usually included in the surprisal measures. This study shows that correcting the surprisal calculation improves n-gram surprisal and that upcoming n-grams affect reading times, replicating previous findings of how lexical frequencies affect reading times. In contrast, the predictivity of PCFG surprisal does not benefit from the surprisal correction despite the fact that lexical sequences skipped by saccades are processed by readers, as demonstrated by the corrected n-gram measure. These results raise questions about the formulation of information-theoretic measures of syntactic processing such as PCFG surprisal and entropy reduction when applied to reading times.
Studies on the role of memory as a predictor of reading time latencies (1) differ in their predictions about when memory effects should occur in processing and (2) have had mixed results, with strong positive effects emerging from isolated constructed stimuli and weak or even negative effects emerging from naturally-occurring stimuli. Our study addresses these concerns by comparing several implementations of prominent sentence processing theories on an exploratory corpus and evaluating the most successful of these on a confirmatory corpus, using a new self-paced reading corpus of seemingly natural narratives constructed to contain an unusually high proportion of memory-intensive constructions. We show highly significant and complementary broad-coverage latency effects both for predictors based on the Dependency Locality Theory and for predictors based on a left-corner parsing model of sentence processing. Our results indicate that memory access during sentence processing does take time, but suggest that stimuli requiring many memory access events may be necessary in order to observe the effect.
Research in computational linguistics, computer graphics and autonomous agents has led to the development of increasingly sophisticated communicative agents over the past few years, bringing new perspective to machine translation research. The engineering of language- based smooth, expressive, natural-looking human gestures can give us useful insights into the design principles that have evolved in natural communication between people. In this paper we prototype a machine translation system from English to American Sign Language (ASL), taking into account not only linguistic but also visual and spatial information associated with ASL signs.