Deriving lexical and syntactic expectation-based measures for psycholinguistic modeling via incremental top-down parsing

Roark, Brian; Bachrach, Asaf; Cárdenas, Carlos; Pallier, Christophe

doi:10.3115/1699510.1699553

Cited by 125 publications

(184 citation statements)

References 34 publications

(55 reference statements)

Supporting

Mentioning

183

Contrasting

Order By: Relevance

“…In contrast, it is less clear exactly how reading times relate to the lexicalized syntactic surprisal metric we used here. Roark et al (2009) reported a significant effect of this type of surprisal on selfpaced reading times, but only for content words once function words had been removed from their analysis. Importantly, however, Roark et al used self-paced reading, which is slower and tends to index later and more strategic language processes compared to eyetracking.…”

Section: Discussionmentioning

confidence: 95%

“…Log-transformed lexical frequency counts were generated from the SUBTLEXus corpus (Brysbaert and New, 2009). Each sentence was parsed using Roark's (Roark, 2001;Roark et al, 2009) incremental top-down PCFG parser trained on the Wall Street Journal corpus of the Penn Treebank (Marcus et al, 1993) to generate syntactic surprisal values for each word.…”

Section: Language Statisticsmentioning

confidence: 99%

“…Roark et al developed an incremental top-down parser (Roark, 2001;Roark et al, 2009) that separates total surprisal into lexical surprisal and syntactic surprisal using an algorithm that builds sets of partial derivations and weights them according to a probabilistic context-free grammar (PCFG). In this parser, total surprisal is decomposed into portions associated with building lexical terminal items (i.e., words) in the parse tree (lexical surprisal) and syntactic structure associated with building nonterminal syntactic nodes (syntactic surprisal).…”

Section: Introductionmentioning

confidence: 99%

“…In this parser, total surprisal is decomposed into portions associated with building lexical terminal items (i.e., words) in the parse tree (lexical surprisal) and syntactic structure associated with building nonterminal syntactic nodes (syntactic surprisal). In an empirical validation of these metrics, Roark et al (2009) showed that self-paced reading times were more accurately predicted from separate estimates of lexical and syntactic surprisal than from a surprisal metric that combined them. Importantly for our purpose of isolating syntactic difficulty, syntactic surprisal in the Roark et al study was uncorrelated with both lexical surprisal and lexical frequency.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Language structure in the brain: A fixation-related fMRI study of syntactic surprisal in reading

et al. 2016

View full text Add to dashboard Cite

How is syntactic analysis implemented by the human brain during language comprehension? The current study combined methods from computational linguistics, eyetracking, and fMRI to address this question. Subjects read passages of text presented as paragraphs while their eye movements were recorded in an MRI scanner. We parsed the text using a probabilistic context-free grammar to isolate syntactic difficulty. Syntactic difficulty was quantified as syntactic surprisal, which is related to the expectedness of a given word's syntactic category given its preceding context. We compared words with high and low syntactic surprisal values that were equated for length, frequency, and lexical surprisal, and used fixation-related (FIRE) fMRI to measure neural activity associated with syntactic surprisal for each fixated word. We observed greater neural activity for high than low syntactic surprisal in two predicted cortical regions previously identified with syntax: left inferior frontal gyrus (IFG) and less robustly, left anterior superior temporal lobe (ATL). These results support the hypothesis that left IFG and ATL play a central role in syntactic analysis during language comprehension. More generally, the results suggest a broader cortical network associated with syntactic prediction that includes increased activity in bilateral IFG and insula, as well as fusiform and right lingual gyri.© 2016 Elsevier Inc. All rights reserved. Keywords:Language Syntax Surprisal Reading Eye movements fMRI IntroductionA complete theory of the human brain must include a description of the neural networks responsible for language processing. Comprehension of connected sentences requires the ability to retrieve words from the lexicon and relate those words to each other, as the input is encountered in real time. The system for creating these groupings is known as syntax, and is generally thought of as a set of computations specifying syntactic categories such as noun and verb, and stating how those categories combine into increasingly larger constituents such as noun phrases, verb phrases, and clauses. The constituent structure of a sentence is the frame on which interpretations are built; the same words will have radically different meanings depending on that structure (e.g., the dog bit the man vs. the man bit the dog). Thus, the human ability to understand language is based both on lexical knowledge and syntactic computations.In this study, we focused on understanding how the human brain implements the syntactic component of the human language faculty. Specifically, we were interested in the nature of the cortical systems that compute syntactic representations during online language comprehension. Three regions have traditionally been identified as candidates for syntactic processing (Fedorenko et al., 2012;Friederici and Kotz, 2003;Grodzinsky and Friederici, 2006;Kaan and Swaab, 2002). Historically, Broca's area or left inferior frontal gyrus (IFG) has been most associated with syntax (Ben-Shachar et al., 2003;Ben-Shachar et al., 2004;...

show abstract

Section: Discussionmentioning

confidence: 95%

Section: Language Statisticsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Language structure in the brain: A fixation-related fMRI study of syntactic surprisal in reading

et al. 2016

View full text Add to dashboard Cite

show abstract

“…Syntactic surprisal and lexical surprisal are calculated to account for high surprisal scores (Roark et al, 2009). As Roark et al (2009) mentions, a word may surprise because it is unconventional, or because it occurs in an unusual context.…”

Section: Methodsmentioning

confidence: 99%

Using Gaze to Predict Text Readability

González-Garduño¹,

Søgaard²

2017

Proceedings of the 12th Workshop on Innovative Use of NLP for Building Educational Applications

View full text Add to dashboard Cite

We show that text readability prediction improves significantly from hard parameter sharing with models predicting first pass duration, total fixation duration and regression duration. Specifically, we induce multi-task Multilayer Perceptrons and Logistic Regression models over sentence representations that capture various aggregate statistics, from two different text readability corpora for English, as well as the Dundee eye-tracking corpus. Our approach leads to significant improvements over Single task learning and over previous systems. In addition, our improvements are consistent across train sample sizes, making our approach especially applicable to small datasets.

show abstract

Sentence Difficulty Analysis with Local Feature Space and Global Distributional Difference

Kim

2012

Convergence and Hybrid Information Technology

View full text Add to dashboard Cite

Abstract. In this paper, we consider the problem of sentence difficulty analysis from various angles. Past works have endeavored to design deterministic scoring algorithms depending only on semantic and syntactic information. We propose instead not only to hire local feature space representing individual sentence with its syntactic and semantic structure, but also to consider global distributional difference among corpora. For the local feature space, we select 28 linguistic features and transform them into conjuncted and discretized form. By applying global score classification, we can show its much improved results. We test our proposed model to 1,000 sentences and get much higher accuracy than traditional learning models such as SVM and AdaBoost.

show abstract

Deriving lexical and syntactic expectation-based measures for psycholinguistic modeling via incremental top-down parsing

Cited by 125 publications

References 34 publications

Language structure in the brain: A fixation-related fMRI study of syntactic surprisal in reading

Language structure in the brain: A fixation-related fMRI study of syntactic surprisal in reading

Using Gaze to Predict Text Readability

Sentence Difficulty Analysis with Local Feature Space and Global Distributional Difference

Contact Info

Product

Resources

About