Towards a Statistical Model of Grammaticality

Alexander Clark; Gianluca Giorgolo; Shalom Lappin

Towards a Statistical Model of Grammaticality

Alexander Clark, Gianluca Giorgolo, Shalom Lappin

Philosophy

Research output: Chapter in Book/Report/Conference proceeding › Poster abstract › peer-review

7 Citations (Scopus)

Abstract

The question of whether it is possible to characterise grammatical
knowledge in probabilistic terms is central to determining
the relationship of linguistic representation to other cognitive
domains. We present a statistical model of grammaticality
which maps the probabilities of a statistical model for
sentences in parts of the British National Corpus (BNC) into
grammaticality scores, using various functions of the parameters
of the model. We test this approach with a classifier on test
sets containing different levels of syntactic infelicity. With appropriate
tuning, the classifiers achieve encouraging levels of
accuracy. These experiments suggest that it may be possible to
characterise grammaticality judgements in probabilistic terms
using an enriched language model.

Original language	English
Title of host publication	Proceedings of the 35th Annual Conference of the Cognitive Science Society 2013
Place of Publication	Berlin
Publisher	Cognitive Science Society.
Pages	2064-2069
Number of pages	6
ISBN (Print)	978-0-9768318-9-1
Publication status	Published - 31 Jul 2013

Publication series

Name	Proceeding of the Annual Conference of the Cognitive Science Society

Keywords

language models, grammaticality, statistical model of grammaticality

Access to Document

http://www.dcs.kcl.ac.uk/staff/lappin/papers/cgl_cogsci13.pdf

Cite this

@inbook{9d10d65e0560490bbb656e58cc2b831b,

title = "Towards a Statistical Model of Grammaticality",

abstract = "The question of whether it is possible to characterise grammaticalknowledge in probabilistic terms is central to determiningthe relationship of linguistic representation to other cognitivedomains. We present a statistical model of grammaticalitywhich maps the probabilities of a statistical model forsentences in parts of the British National Corpus (BNC) intogrammaticality scores, using various functions of the parametersof the model. We test this approach with a classifier on testsets containing different levels of syntactic infelicity. With appropriatetuning, the classifiers achieve encouraging levels ofaccuracy. These experiments suggest that it may be possible tocharacterise grammaticality judgements in probabilistic termsusing an enriched language model.",

keywords = "language models, grammaticality, statistical model of grammaticality",

author = "Alexander Clark and Gianluca Giorgolo and Shalom Lappin",

year = "2013",

month = jul,

day = "31",

language = "English",

isbn = "978-0-9768318-9-1",

series = "Proceeding of the Annual Conference of the Cognitive Science Society",

publisher = "Cognitive Science Society. ",

pages = "2064--2069",

booktitle = "Proceedings of the 35th Annual Conference of the Cognitive Science Society 2013",

}

Towards a Statistical Model of Grammaticality. / Clark, Alexander ; Giorgolo, Gianluca ; Lappin, Shalom.
Proceedings of the 35th Annual Conference of the Cognitive Science Society 2013. Berlin: Cognitive Science Society. , 2013. p. 2064-2069 (Proceeding of the Annual Conference of the Cognitive Science Society).

Research output: Chapter in Book/Report/Conference proceeding › Poster abstract › peer-review

TY - CHAP

T1 - Towards a Statistical Model of Grammaticality

AU - Clark, Alexander

AU - Giorgolo, Gianluca

AU - Lappin, Shalom

PY - 2013/7/31

Y1 - 2013/7/31

N2 - The question of whether it is possible to characterise grammaticalknowledge in probabilistic terms is central to determiningthe relationship of linguistic representation to other cognitivedomains. We present a statistical model of grammaticalitywhich maps the probabilities of a statistical model forsentences in parts of the British National Corpus (BNC) intogrammaticality scores, using various functions of the parametersof the model. We test this approach with a classifier on testsets containing different levels of syntactic infelicity. With appropriatetuning, the classifiers achieve encouraging levels ofaccuracy. These experiments suggest that it may be possible tocharacterise grammaticality judgements in probabilistic termsusing an enriched language model.

AB - The question of whether it is possible to characterise grammaticalknowledge in probabilistic terms is central to determiningthe relationship of linguistic representation to other cognitivedomains. We present a statistical model of grammaticalitywhich maps the probabilities of a statistical model forsentences in parts of the British National Corpus (BNC) intogrammaticality scores, using various functions of the parametersof the model. We test this approach with a classifier on testsets containing different levels of syntactic infelicity. With appropriatetuning, the classifiers achieve encouraging levels ofaccuracy. These experiments suggest that it may be possible tocharacterise grammaticality judgements in probabilistic termsusing an enriched language model.

KW - language models, grammaticality, statistical model of grammaticality

M3 - Poster abstract

SN - 978-0-9768318-9-1

T3 - Proceeding of the Annual Conference of the Cognitive Science Society

SP - 2064

EP - 2069

BT - Proceedings of the 35th Annual Conference of the Cognitive Science Society 2013

PB - Cognitive Science Society.

CY - Berlin

ER -

Towards a Statistical Model of Grammaticality

Abstract

Publication series

Keywords

Access to Document

Fingerprint

Cite this