"Bridging the gap between subsymbolic and symbolic techniques: A pragmatic approach"

G. Stamou, D.Vogiatzis (Greece) and S. Strove (The Netherlands)

Abstract: We present a survey of methods for passing from subsymbols to symbols. First, we present a definition of the following notions: Subsymbols, Symbols and Hybrid Systems, as well as a taxonomy of Intelligent Systems. Then, pure Artificial Intelligent (AI) as well as pure Computational Intelligence (CI) techniques for subsymbolic to symbolic mapping are exposed. Next, we present a taxonomy of hybrid systems; according to that there are three categories of hybrid systems: Combined Intelligent Systems, Transformational Intelligent Systems and Coupled Intelligent Systems. Hybrid systems are a very promising alternative to pure AI and pure CI techniques for mapping subsymbols to symbols. Finally, we elaborate on the Hybrid Intelligent Systems as well as on the approach, which we follow in Physta research project, which is a subcategory of Hybrid Systems.

Key-Words: Neurosymbolic Integration, Neural Networks, Symbols