How Language Processing can Shape a Common Model of Cognition

Research output: Contribution to journalConference article

Abstract

What role does the study of natural language play in the task of developing a unified theory and common model of cognition? Language is perhaps the most complex behaviour that humans exhibit, and, as such, is one of the most difficult problems for understanding human cognition. Linguistic theory can both inform and be informed by unified models of cognition. We discuss (1) how computational models of human cognition can provide insight into how humans produce and comprehend language and (2) how the problem of modelling language processing raises questions and creates challenges for widely used computational models of cognition. Evidence from the literature suggests that behavioural phenomena, such as recency and priming effects, and cognitive constraints, such as working memory limits, affect how language is produced by humans in ways that can be predicted by computational cognitive models. But just as computational models can provide new insights into language, language can serve as a test for these models. For example, simulating language learning requires the use of more powerful machine learning techniques, such as deep learning and vector symbolic architectures, and language comprehension requires a capacity for on-the-fly situational model construction. In sum, language plays an important role in both shaping the development of a common model of the mind, and, in turn, the theoretical understanding of language stands to benefit greatly from the development of a common model.

Original languageEnglish (US)
Pages (from-to)724-729
Number of pages6
JournalProcedia Computer Science
Volume145
DOIs
StatePublished - Jan 1 2018
Event9th Annual International Conference on Biologically Inspired Cognitive Architectures, BICA 2018 - Prague, Czech Republic
Duration: Aug 22 2018Aug 24 2018

Fingerprint

Processing
Linguistics
Learning systems
Data storage equipment

All Science Journal Classification (ASJC) codes

  • Computer Science(all)

Cite this

@article{097994e9e8244091ab80233b78a7cd11,
title = "How Language Processing can Shape a Common Model of Cognition",
abstract = "What role does the study of natural language play in the task of developing a unified theory and common model of cognition? Language is perhaps the most complex behaviour that humans exhibit, and, as such, is one of the most difficult problems for understanding human cognition. Linguistic theory can both inform and be informed by unified models of cognition. We discuss (1) how computational models of human cognition can provide insight into how humans produce and comprehend language and (2) how the problem of modelling language processing raises questions and creates challenges for widely used computational models of cognition. Evidence from the literature suggests that behavioural phenomena, such as recency and priming effects, and cognitive constraints, such as working memory limits, affect how language is produced by humans in ways that can be predicted by computational cognitive models. But just as computational models can provide new insights into language, language can serve as a test for these models. For example, simulating language learning requires the use of more powerful machine learning techniques, such as deep learning and vector symbolic architectures, and language comprehension requires a capacity for on-the-fly situational model construction. In sum, language plays an important role in both shaping the development of a common model of the mind, and, in turn, the theoretical understanding of language stands to benefit greatly from the development of a common model.",
author = "Matthew Kelly and Reitter, {David T.}",
year = "2018",
month = "1",
day = "1",
doi = "10.1016/j.procs.2018.11.047",
language = "English (US)",
volume = "145",
pages = "724--729",
journal = "Procedia Computer Science",
issn = "1877-0509",
publisher = "Elsevier BV",

}

How Language Processing can Shape a Common Model of Cognition. / Kelly, Matthew; Reitter, David T.

In: Procedia Computer Science, Vol. 145, 01.01.2018, p. 724-729.

Research output: Contribution to journalConference article

TY - JOUR

T1 - How Language Processing can Shape a Common Model of Cognition

AU - Kelly, Matthew

AU - Reitter, David T.

PY - 2018/1/1

Y1 - 2018/1/1

N2 - What role does the study of natural language play in the task of developing a unified theory and common model of cognition? Language is perhaps the most complex behaviour that humans exhibit, and, as such, is one of the most difficult problems for understanding human cognition. Linguistic theory can both inform and be informed by unified models of cognition. We discuss (1) how computational models of human cognition can provide insight into how humans produce and comprehend language and (2) how the problem of modelling language processing raises questions and creates challenges for widely used computational models of cognition. Evidence from the literature suggests that behavioural phenomena, such as recency and priming effects, and cognitive constraints, such as working memory limits, affect how language is produced by humans in ways that can be predicted by computational cognitive models. But just as computational models can provide new insights into language, language can serve as a test for these models. For example, simulating language learning requires the use of more powerful machine learning techniques, such as deep learning and vector symbolic architectures, and language comprehension requires a capacity for on-the-fly situational model construction. In sum, language plays an important role in both shaping the development of a common model of the mind, and, in turn, the theoretical understanding of language stands to benefit greatly from the development of a common model.

AB - What role does the study of natural language play in the task of developing a unified theory and common model of cognition? Language is perhaps the most complex behaviour that humans exhibit, and, as such, is one of the most difficult problems for understanding human cognition. Linguistic theory can both inform and be informed by unified models of cognition. We discuss (1) how computational models of human cognition can provide insight into how humans produce and comprehend language and (2) how the problem of modelling language processing raises questions and creates challenges for widely used computational models of cognition. Evidence from the literature suggests that behavioural phenomena, such as recency and priming effects, and cognitive constraints, such as working memory limits, affect how language is produced by humans in ways that can be predicted by computational cognitive models. But just as computational models can provide new insights into language, language can serve as a test for these models. For example, simulating language learning requires the use of more powerful machine learning techniques, such as deep learning and vector symbolic architectures, and language comprehension requires a capacity for on-the-fly situational model construction. In sum, language plays an important role in both shaping the development of a common model of the mind, and, in turn, the theoretical understanding of language stands to benefit greatly from the development of a common model.

UR - http://www.scopus.com/inward/record.url?scp=85059460790&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85059460790&partnerID=8YFLogxK

U2 - 10.1016/j.procs.2018.11.047

DO - 10.1016/j.procs.2018.11.047

M3 - Conference article

VL - 145

SP - 724

EP - 729

JO - Procedia Computer Science

JF - Procedia Computer Science

SN - 1877-0509

ER -