Abstract:

In this article we study the use of self-organizing neural networks (SOM - Self Organizing Map) to categorize concepts and relationships in order to generate and edit ontologies autonomously. The first point to consider is how to characterize various concepts and relationships for a self-organizing neural network can to categorize them appropriately. And the next consideration is to find the algorithm that recognizes the different aspects to the functioning of the neural network and found at these concepts and relationships to describe the ontology generated by the neural network.

Ontologies are a mechanism developed by technology. These are made in order to make domain knowledge representations that humans derive naturally from the environment abstractions. Currently these representations are built with the help of experts in the area of knowledge about the ontology is developed and a computer expert to translate this information into concepts and relations with sense to a computer.

However this approach is working and underestimate the power of ontologies trying them static taxonomies of concepts and relationships. The current area of research is a different approach to ontologies. This research Seeks to provide autonomy ontology to grow dynamically according to the needs and purposes of its creation.

This article intends provide an alternative for the generation of these ontologies in a dynamic / autonomous, making ontologies learn the environment and input stimuli by using a type of unsupervised neural network called self-organizing maps.

Keywords:

SOM, Self Organizing Map, ontology, autonomous, dynamic, concept, relation.