Optimizing the Multiling model for word translagion: comprehension versus production

Abstract

In this thesis, it is described how the Multilink model for word translation (Dijkstra & Rekké, 2010) was altered in an attempt to better account for lexical effects found in empirical linguistic data, such as cognate status and word frequency. An analysis was done to determine whether changes to the Multilink model had improved it. This was done by splitting the translation task into different components, namely a comprehension and a production component. Data simulated with the new version of Multilink was found to show high correlations with empirical data for comprehension and moderate correlations with empirical data for production. It was also found that Multilink does still not correctly account for cognate status and word frequency. Despite this, simulated translation data shows high correlations with empirical data, making Multilink an excellent candidate for further research.