Parsing as a Cue‐Based Retrieval Model

Abstract

This paper develops a novel psycholinguistic parser and tests it against experimental and corpus reading data. The parser builds on the recent research into memory structures, which argues that memory retrieval is content-addressable and cue-based. It is shown that the theory of cue-based memory systems can be combined with transition-based parsing to produce a parser that, when combined with the cognitive architecture ACT-R, can model reading and predict online behavioral measures (reading times and regressions). The parser's modeling capacities are tested against self-paced reading experimental data (Grodner & Gibson, 2005), eye-tracking experimental data (Staub, 2011), and a self-paced reading corpus (Futrell et al., 2018).

Robust Effects of Working Memory Demand during Naturalistic Language Comprehension in Language-Selective Cortex

“…These exploratory results have several implications. First, they support the existence of syntacticallyrelated WM load in the language network during naturalistic sentence comprehension and indicate that the DLT captures this WM signature better than more recent algorithmic-level models of WM in language (Lewis & Vasishth, 2005;Rasmussen & Schuler, 2018;Dotlačil, 2021). Second, they present a serious challenge to the hypothesis that the MD networkthe most likely domain-general WM resourceis recruited for language-related WM operations: despite casting a broad net over theoretically motivated WM measures and performing the testing in-sample, the MD network does not show systematic correlates of WM demand.…”

“…At a high level, we analyze the influence of theory-driven measures of working memory load during auditory comprehension of naturalistic stories (Futrell et al, 2020) on activation levels in the language-selective (LANG) vs. domain-general multiple-demand (MD) networks identified in each participant using an independent functional localizer. To control for regional and participant-level variation in the hemodynamic response function (HRF), the HRF is estimated from data using continuous-time deconvolutional regression (Shain & Schuler, 2018, 2021, rather than assumed (cf. e.g., Bhattasali et al, 2019;Brennan et al, 2016).…”

“…This implementation does not cover arbitrary sentences of English and cannot therefore be applied to our stimuli without extensive additional engineering of the parsing rules. However, a recently proposed modification to the ACT-R framework has a broad-coverage implementation and has already been applied to model reading time responses to the same set of stories (Dotlačil, 2021). It does so by moving the parsing rules from procedural to declarative memory, allowing the rules themselves to be retrieved and activated in the same manner as parse fragments.…”

“…ACT-R target activation is expected to be anticorrelated with retrieval difficulty. See Dotlačil (2021) and Lewis & Vasishth (2005) for details.…”

SEAM: An Integrated Activation-Coupled Model of Sentence Processing and Eye Movements in Reading

“…These models focus on different aspects of sentence processing, and have been evaluated against corpus data, such as the Schilling corpus (Schilling et al, 1998). Two models that investigate the interaction between eye-movement control and sentence comprehension using data from planned experiments are reported in Vasishth and Engelmann (2022) and Dotlačil (2021); both these investigations use a highly simplified version of E-Z Reader, that is, the Eye Movements and Movement of Attention (EMMA) model embedded within the ACT-R architecture (Salvucci, 2001). The simplified EMMA model has important limitations; for example, as discussed in Engelmann et al (2013), the model only allows regressive eye movements to the preceding word.…”