WORDINFO

How do words inform? Explaining the role of information theory in language comprehension

Coordinatore	STICHTING KATHOLIEKE UNIVERSITEIT    more  Organization address address: GEERT GROOTEPLEIN NOORD 9 city: NIJMEGEN postcode: 6525 EZ contact info Titolo: Mrs. Nome: Laura Cognome: Pander Email: send email Telefono: +31 24 361 2135 Fax: +31 24 3615481
Nazionalità Coordinatore	Netherlands [NL]
Totale costo	75˙000 €
EC contributo	75˙000 €
Programma	FP7-PEOPLE Specific programme "People" implementing the Seventh Framework Programme of the European Community for research, technological development and demonstration activities (2007 to 2013)
Code Call	FP7-PEOPLE-2012-CIG
Funding Scheme	MC-CIG
Anno di inizio	2013
Periodo (anno-mese-giorno)	2013-03-01   -   2016-02-29

 Partecipanti

#	participant	country	role	EC contrib. [€]
1	STICHTING KATHOLIEKE UNIVERSITEIT  Organization address address: GEERT GROOTEPLEIN NOORD 9 city: NIJMEGEN postcode: 6525 EZ contact info Titolo: Mrs. Nome: Laura Cognome: Pander Email: send email Telefono: +31 24 361 2135 Fax: +31 24 3615481	NL (NIJMEGEN)	coordinator	75˙000.00

Mappa

 Word cloud

Esplora la "nuvola delle parole (Word Cloud) per avere un'idea di massima del progetto.

 Obiettivo del progetto (Objective)

'Cognition is often said to arise from information processing. Indeed, information theory has been applied to explain aspects of human cognition, such as fluctuations in cognitive load during reading: By computing how much information each word conveys, it can be shown that reading more informative words takes longer, and leads to dilation of the reader’s pupils. However, it is as yet unclear how these observed effects of word information come about: What are the underlying mechanisms from which they emerge? The proposed project aims to fill this gap in our understanding of human language, by combing computational modelling and human experimental research. First, an original model is developed, based on the Deep Belief Network (DBN) architecture, that extracts language statistics from a large text corpus. As a well-defined probability model, the DBN can compute word-information values over a set of test sentences. These values, which depend only on the acquired language statistics, form _predictions_ of word-reading times and pupil size. In addition, DBNs are neurally inspired and can be made to process input dynamically, mimicking cognitive processing over continuous time. This yields _simulated_ reading times, which depend on the model’s processing assumptions. Likewise, model-internal dynamics will be taken to simulate cognitive load that correlates with pupil size. Next, in eye-tracking and pupillometry studies, _actual_ reading-time and pupil-size data are collected over the same test sentences as processed by the model. Statistical analyses of the relations between predicted, simulated, and actual data will reveal how, exactly, the effects of word information arise. This greatly increases our understanding of the structures and mechanisms involved in language comprehension. Moreover, the project will yield a more accurate cognitive interpretation of pupil dilation and can elucidate the relation between information-theoretical and psychological constructs.'