skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: Why Does Surprisal From Larger Transformer-Based Language Models Provide a Poorer Fit to Human Reading Times?
Abstract 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.  more » « less
Award ID(s):
1816891
PAR ID:
10417899
Author(s) / Creator(s):
;
Date Published:
Journal Name:
Transactions of the Association for Computational Linguistics
Volume:
11
ISSN:
2307-387X
Page Range / eLocation ID:
336 to 350
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; ; (, Transactions of the Association for Computational Linguistics)
    Abstract Surprisal theory posits that less-predictable words should take more time to process, with word predictability quantified as surprisal, i.e., negative log probability in context. While evidence supporting the predictions of surprisal theory has been replicated widely, much of it has focused on a very narrow slice of data: native English speakers reading English texts. Indeed, no comprehensive multilingual analysis exists. We address this gap in the current literature by investigating the relationship between surprisal and reading times in eleven different languages, distributed across five language families. Deriving estimates from language models trained on monolingual and multilingual corpora, we test three predictions associated with surprisal theory: (i) whether surprisal is predictive of reading times, (ii) whether expected surprisal, i.e., contextual entropy, is predictive of reading times, and (iii) whether the linking function between surprisal and reading times is linear. We find that all three predictions are borne out crosslinguistically. By focusing on a more diverse set of languages, we argue that these results offer the most robust link to date between information theory and incremental language processing across languages. 
    more » « less
  2. ; ; (, Frontiers in Artificial Intelligence)
    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 article first describes an incremental left-corner parser that incorporates information about common linguistic abstractions such as syntactic categories, predicate-argument structure, and morphological rules as a computational-level model of sentence processing. The article then evaluates a variety of structural parsers and deep neural language models as cognitive models of sentence processing by comparing the predictive power of their surprisal estimates on self-paced reading, eye-tracking, and fMRI data collected during real-time language processing. The results show that surprisal estimates from the proposed left-corner processing model deliver comparable and often superior fits to self-paced reading and eye-tracking data when compared to those from neural language models trained on much more data. This may suggest that the strong linguistic generalizations made by the proposed processing model may help predict humanlike processing costs that manifest in latency-based measures, even when the amount of training data is limited. Additionally, experiments using Transformer-based language models sharing the same primary architecture and training data show a surprising negative correlation between parameter count and fit to self-paced reading and eye-tracking data. These findings suggest that large-scale neural language models are making weaker generalizations based on patterns of lexical items rather than stronger, more humanlike generalizations based on linguistic structure. 
    more » « less
  3. ; ; ; ; (, Cognitive Science)
    Abstract Source code is a form of human communication, albeit one where the information shared between the programmers reading and writing the code is constrained by the requirement that the code executes correctly. Programming languages are more syntactically constrained than natural languages, but they are also very expressive, allowing a great many different ways to express even very simple computations. Still, code written by developers is highly predictable, and many programming tools have taken advantage of this phenomenon, relying on language modelsurprisalas a guiding mechanism. While surprisal has been validated as a measure of cognitive load in natural language, its relation to human cognitive processes in code is still poorly understood. In this paper, we explore the relationship between surprisal and programmer preference at a small granularity—do programmers prefer more predictable expressions in code? Usingmeaning‐preserving transformations, we produce equivalent alternatives to developer‐written code expressions and run a corpus study on Java and Python projects. In general, language models rate the code expressions developerschooseto write as more predictable than these transformed alternatives. Then, we perform two human subject studies asking participants to choose between two equivalent snippets of Java code with different surprisal scores (one original and transformed). We find that programmersdoprefer more predictable variants, and that stronger language models like the transformer align more often and more consistently with these preferences. 
    more » « less
  4. ; ; ; ; ; ; ; ; (, Machine Learning: Science and Technology)
    Abstract Pre-trained transformer language models (LMs) on large unlabeled corpus have produced state-of-the-art results in natural language processing, organic molecule design, and protein sequence generation. However, no such models have been applied to learn the composition patterns for the generative design of material compositions. Here we train a series of seven modern transformer models (GPT, GPT-2, GPT-Neo, GPT-J, BLMM, BART, and RoBERTa) for materials design using the expanded formulas of the ICSD, OQMD, and Materials Projects databases. Six different datasets with/out non-charge-neutral or EB samples are used to benchmark the generative design performances and uncover the biases of modern transformer models for the generative design of materials compositions. Our experiments show that the materials transformers based on causal LMs can generate chemically valid material compositions with as high as 97.61% to be charge neutral and 91.22% to be electronegativity balanced, which has more than six times higher enrichment compared to the baseline pseudo-random sampling algorithm. Our LMs also demonstrate high generation novelty and their potential in new materials discovery is proved by their capability to recover the leave-out materials. We also find that the properties of the generated compositions can be tailored by training the models with selected training sets such as high-bandgap samples. Our experiments also show that different models each have their own preference in terms of the properties of the generated samples and their running time complexity varies a lot. We have applied our materials transformers to discover a set of new materials as validated using density functional theory calculations. All our trained materials transformer models and code can be accessed freely at http://www.github.com/usccolumbia/MTransformer . 
    more » « less
  5. ; ; (, Findings of the Association for Computational Linguistics: EACL)
    Yvette Graham, Matthew Purver (Ed.)
    Understanding the dynamics of counseling conversations is an important task, yet it is a challenging NLP problem regardless of the recent advance of Transformer-based pre-trained language models. This paper proposes a systematic approach to examine the efficacy of domain knowledge and large language models (LLMs) in better representing conversations between a crisis counselor and a help seeker. We empirically show that state-of-the-art language models such as Transformer-based models and GPT models fail to predict the conversation outcome. To provide richer context to conversations, we incorporate human-annotated domain knowledge and LLM-generated features; simple integration of domain knowledge and LLM features improves the model performance by approximately 15%. We argue that both domain knowledge and LLM-generated features can be exploited to better characterize counseling conversations when they are used as an additional context to conversations. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email