Organizing conceptual graphs for fast knowledge retrieval

Abstract

The expressiveness of conceptual graphs allows for accurate representation of knowledge in the domain of information retrieval. However, the complexity of the operators introduced by this formalism are in contradiction with the necessary speed of a practicable information retrieval system. In particular, the projection operator is considered to be time-consuming. We propose a new logic-based organization of conceptual graphs. Considering conceptual graphs from a logical point of view leads to a form of conceptual graphs that we call standard logical form. In this case, the complexity of our retrieval algorithm is polynomial. Retrieval exploits the logic-based organization, and the iterative application of the projection operator in each graph of the collection is changed by a single application of an algorithm to the whole collection. The algorithm is presented in detail and its complexity is studied. Experimentation on a collection of 650 images, by using an already existent conceptual graph platform, shows net improvement in retrieval time performance.