Models for the Categorization of Natural Language Concepts
Speaker: Trevor Fountain
Abstract:
In this talk I present work toward building a model of categorization for natural language data. I begin by explaining the nature of categorization and detailing why current cognitive models are unable to handle natural language; I then outline how these models might be adapted to such a task, and present results evaluating their performance on a number of categorization-related tasks. In order to perform these evaluations it is necessary to collect human data for the same tasks, therefore I also describe a significant data collection experiment designed to collect this information. I conclude by discussing why the adapted models are poorly suited to the task of categorizing natural language data, and by indicating likely directions for ways in which they might be improved.
