Optimizing the Multilink model for word translation: A simulation study to improve model-to-data fit
Keywords
No Thumbnail Available
Authors
Issue Date
2016-07-13
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
In this study, the computational Multilink model (Dijkstra and
Rekké, 2010) is used to simulate reaction times of word translation
production data. The Multilink model aims amongst others to simulate
the word translation process by accounting for effects of cognate status,
word frequency, and translation direction. Earlier studies (e.g., Peacock,
2015) have shown that Multilink is able to capture the differences in
reaction time difference quite well. In this thesis, the Multilink model is
further optimized via a number of modifications. The effect of the
modifications is evaluated with a fitness measure that incorporates both
reaction time differences as well as translation accuracy and is compared
to experimental reaction time data collected by Pruijn (2015). The
modifications lead to a 28.4% in fitness and eliminated all translation
errors. The model was improved in particular by 1) Improving effects of
word frequency through the resting level activation of orthographic and
phonological nodes and 2) Modifications on the activation function from
orthographic to semantic nodes. Altogether, the optimized Multilink
model provides a solid framework for further research upon the effects
that occur during word translation.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen