Retrieving Word Associations with a Simple Neighborhood Algorithm in a Graph-Based Resource

Abstract

The paper explains the procedure to obtain word associations starting from a graph that has not been specifically built for that purpose. Our goal is being able to simulate human word associations by using the simplest possible methods, including the basic tools of a co-occurrence network from a non-annotated corpus, and a very simple search algorithm based on neighborhood. The method has been tested in the Cogalex shared task, revealing the difficulty of achieving word associations without semantic annotation.