Tilburg center for Cognition and Communication (TiCC)

We study how people communicate with each other and how computer systems can be taught to communicate with us.

vignetweb

TiCC Colloquium: Simon De Deyne & Gert Storms

What: Using large-scale empirical semantic networks to study meaning in the mental lexicon
Where: TZ 3
When: Wednesday, 21 September 2016, 12:30 - 13:30 hours


Abstract

An individual's mental lexicon contains the knowledge about words she or he acquired over a lifetime. One way to study how word meaning is represented in this lexicon is by using a semantic network representation. In this talk, we explain how such a network, where words correspond to nodes which are connected through a variety of semantic relations can be derived from word associations. While representing meaning as a semantic network is central to many cognitive theories, we implement a working model at a realistic scale as it contains the meaning of the majority of words people know.

We will focus on the nature of semantic representations by investigating how a large empirical network is structured at different scales. In the first part of this talk, we discuss a series of studies showing how word meaning is organized globally. At the global scale, our results indicate an organization that distinguishes positive and negative words in line with the seminal work by Osgood and extends by identifying concreteness as one of the main determinants of structure in the mental lexicon. At an intermediate scale, the network indicates that most semantic categories are structured along thematic links rather than taxonomic links.

In the second part of this talk we elaborate further on how meaning is structured within different domains using a series of experiments where participants judge relatedness. We show relatedness can be derived by investigating the overlap between the response distributions of words and propose a dynamic approach to infer meaning using random walks. This allow us to estimate how similar two remotely related words are even if they're not directly related.

In the final part we highlight some applications of this approach where we take a developmental (how networks change over time) and clinical perspective (how networks change in certain disorders).

When: 21 September 2016 12:30

End date: 21 September 2016 13:30