Hybrid Architecture for Web Search Systems Based on Hierarchical Taxonomies

Search systems based on hierarchical taxonomies provide a specific type of search functionality that is not provided by conventional search engines. For instance, using a taxonomy, the user can look for documents related to just one of the categories of the taxonomy. This paper describes a hybrid data architecture that improves the performance of restricted searches for a few categories of a taxonomy. The proposed architecture is based on a hybrid data structure composed of an inverted file with multiple integrated signature files. A detailed analysis of superimposing codes on directed acyclic graphs proves that they adapt perfectly well to a search system based on a hierarchical ontology. Two variants are presented: the hybrid architecture with complete information and the hybrid architecture with partial information. The validity of this hybrid architecture was analyzed by developing and comparing it with a basic architecture. The performance of restricted queries is clearly improved, especially with the hybrid architecture with partial information. This variant outperformed by 50 % the basic architecture for all workload environments, with a slight reduction in performance for the lower levels of the graph.