Table of Contents Author Guidelines Submit a Manuscript
Computational Intelligence and Neuroscience
Volume 2014, Article ID 920892, 11 pages
http://dx.doi.org/10.1155/2014/920892
Research Article

Mental Mechanisms for Topics Identification

Department of Mathematics and Computer Science, Royal Military College, Kingston, ON, Canada K7K 7B4

Received 19 November 2013; Accepted 4 February 2014; Published 13 March 2014

Academic Editor: Jianwei Shuai

Copyright © 2014 Louis Massey. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Topics identification (TI) is the process that consists in determining the main themes present in natural language documents. The current TI modeling paradigm aims at acquiring semantic information from statistic properties of large text datasets. We investigate the mental mechanisms responsible for the identification of topics in a single document given existing knowledge. Our main hypothesis is that topics are the result of accumulated neural activation of loosely organized information stored in long-term memory (LTM). We experimentally tested our hypothesis with a computational model that simulates LTM activation. The model assumes activation decay as an unavoidable phenomenon originating from the bioelectric nature of neural systems. Since decay should negatively affect the quality of topics, the model predicts the presence of short-term memory (STM) to keep the focus of attention on a few words, with the expected outcome of restoring quality to a baseline level. Our experiments measured topics quality of over 300 documents with various decay rates and STM capacity. Our results showed that accumulated activation of loosely organized information was an effective mental computational commodity to identify topics. It was furthermore confirmed that rapid decay is detrimental to topics quality but that limited capacity STM restores quality to a baseline level, even exceeding it slightly.