System recommending novel and semantically linked words

Algorithm based on free energy principle.

Abstract

Novel and semantically associated words facilitate broadening ideas. In this study, we propose a computational system that recommends novel and semantically associated words. The algorithm of the system is based on free energy principle explaining cognition of organisms in a general manner. The system recommends words while controlling a balance between novelty and semantic connection. The system's outputs are compared with those of Word2vec, and its effectiveness is discussed.