Study on Human-Machine Semantics: An Artificial Intelligence Perspective

CHEN Xiao-ping
DOI: https://doi.org/10.3785/j.issn.1008-942X.2006.03.002
2006-01-01
Abstract:Semantic web is one of the main trends in current research on web application. From the point of view of Artificial Intelligence, we investigate the development of semantic web and its related domains in AI such as natural language understanding, knowledge representation and reasoning, and ontology, through which we get some important observations. Firstly, the relationship between human and machine is changing rapidly—— the network of computers becomes the basic bridge in the communication and interaction between human and machine. Secondly, human-machine interaction comes to gain a new feature —— semantic processing will take a more and more important role, while syntactic processing is dominant currently. Thirdly, the formal characteristic of machine evolves from homogeneous system to heterogeneous system. A classical logic system is a homogeneous system in the sense that it provides a unique ontology and a unique inferential mechanism so that soundness and completeness hold. The very notion of soundness and completeness has been accepted as a criterion for ″correctness″ of AI systems in many subfields of AI such as knowledge representation and reasoning, natural language understanding, and planning. However, this criterion is inadequate for the expected ″global system″ generating from semantic web in the future. Instead, ″provable correctness with respect to semantic web rules″ is now considered to be an adequate alternative. The reason is that the ″global system″ will be a heterogeneous system, one equipped with multiple ontologies and multiple inferential mechanisms, where this new criterion is adopted in order to keep the integrality of the global system. The term ″ontology″ is borrowed from philosophy and assigned a new meaning, i.e., ″a formal specification of a conceptualization″. This notion and technique proves extremely important in AI development, especially in very large-scale knowledge systems such as Cyc, which is designed to capture all of the knowledge of an average teenage. The analysis above reveals that AI development along the line of semantic web begins to depart from its ″original sketch″ embodied by Leibniz’s Characteristica Universalis, which can be considered as a universal homogeneous system of human knowledge and communication. The most significant evolvement of AI thought behind this department is the emergence of the possibility of ″human-machine semantics″, a novel semantics shared by human and machine. One can characterize this semantics with concepts of heterogeneity, provable correctness with respect to semantic web rules, understandability to both human and machine.
What problem does this paper attempt to address?