Fusing conceptual graphs and fuzzy logic: towards the structure and expressiveness of natural language

• Authors: Tru H. Cao
• Venue: IUKM'11 Proceedings of the 2011 international conference on Integrated uncertainty in knowledge modelling and decision making
• Year: 2011

Natural language is a principal and important means of human communication. It is used to express information as inputs to be processed by human brains then, very often, outputs are also expressed in natural language. The capacity for humans to communicate using language allows us to give, receive, and understand information expressed within a rich and flexible representational framework. Moreover, we can reason based on natural language expressions, and make decisions based on the information they convey, though this information usually involves imprecise terms and uncertain facts. How humans process information represented in natural language is still a challenge to science, in general, and to Artificial Intelligence, in particular. However, it is clear that, for a computer with the conventional processing paradigm to handle natural language, a formalism is required. For reasoning, it is desirable that such a formalism be a logical one. A logic for handling natural language should have not only a structure of formulas close to that of natural language sentences, but also a capability to deal with the semantics of vague linguistic terms pervasive in natural language expressions. Conceptual graphs (Sowa [2,3]) and fuzzy logic (Zadeh [7,8]) are two logical formalisms that emphasize the target of natural language, each of which is focused on one of the two mentioned desired features of a logic for handling natural language. While a smooth mapping between logic and natural language has been regarded as the main motivation of conceptual graphs (Sowa [4,5,6]), a methodology for computing with words has been regarded as the main contribution of fuzzy logic (Zadeh [9,10,11]). However, although conceptual graphs and fuzzy logic have the common target of natural language, until recently they were studied and developed quite separately. Their combination would be a great advantage towards a knowledge representation language that can approach the structure and expressiveness of natural language. At this juncture, conceptual graphs provide a syntactic structure for a smooth mapping to and from natural language, while fuzzy logic provides a semantic processor for approximate reasoning with words having vague meanings. This talk presents the combined result of an interdisciplinary research programme focused on the integration of conceptual graphs and fuzzy logic, towards a knowledge representation language that is close to natural language in both of the structure and expressiveness (Cao [1]). First, the talk summarizes the development of fuzzy conceptual graphs and their logic programming foundations, as a graph-based order-sorted fuzzy set logic programming language for automated reasoning with fuzzy object attributes and types. Second, it presents the extension of fuzzy conceptual graphs with general quantifiers and direct reasoning operations on these extended conceptual graphs, which could be mapped to and from generally quantified natural language statements. Third, it introduces recent applications of fuzzy conceptual graphs for understanding natural language queries and semantic search.