Automatic Detection of Semantics Primitives in Explicative Dictionaries with Bio-inspired Algorithms

Abstract

Inevitably, any explanatory dictionary contains cycles in its definitions, that is, if a word is defined in the dictionary and then used in a definition, there is always a path in the dictionary that returns to the same word. In a good dictionary the cycles are long, but they are unavoidable. A computational dictionary cannot contain any cycles in its definitions without them affecting the ability of logical inference of computer systems. In this study, we name semantic primitives to such words in the dictionary that if removed, the cycles would be eliminated; that is, those words would not have a definition and, in this sense, they are primitive. In this research, our goal is to keep as many words in the dictionary, i.e., to minimize the number of semantic primitives. We present a method that achieves the smallest set of primitives obtained so far. In order to accomplish this, the representation of the dictionary was used as a directed graph, and a differential evolution algorithm, that determines the order in which the graph should be built, was applied to the dictionary.