Expanding the Multilink lexicon with the French language A collection of simulation studies

Thumbnail Image
Journal Title
Journal ISSN
Volume Title
Multilink is a computational model that simulates human word recognition and translation processes according to a localist-connectionist paradigm. The input word results in the activation of the target word, its orthographic neighbours, and semantically related words. In its current form, Multilink only accommodates bilin- gual Dutch-English word recognition and translation tasks. This thesis describes the extension of the current lexicon with French vocabulary. Furthermore, two simulation studies were performed to measure the performance of Multilink with its newly acquired lexicon. The rst study compared variations of the new French lexicon, such as using dif- ferent frequency measures and the presence or absence of diacritical characters. These variations were tested by Multilink simulations of the lexical decision task. Simulation data (cycle times) were correlated with reaction times derived from the French Lexicon Project (Ferrand et al., 2010). A lexicon involving movie-derived frequencies produced a model-to-data correlation of 0.391 (n=832). The presence of diacritical characters did not produce a notable di erence in correlation. The second study simulated the empirical study by Peeters, Dijkstra, and Grainger (2013). This study examined the cognate facilitation e ect by performing English- French lexical decision tasks. The results of this simulation produced a model-to- data correlation of 0.667 (n=166, p < 0.0001). These results show that Multilink is capable of simulating word recognition pro- cesses involving the French language. Especially in the French-English bilingual domain, Multilink scores very well.
Faculteit der Sociale Wetenschappen