Adding Knowledge Extracted by Association Rules into Similarity Queries

Mônica R. P. Ferreira, Marcela X Ribeiro, Agma J. M. Traina, Richard Chbeir, Caetano Traina Jr.


In this paper, we propose new techniques to improve the quality of similarity queries over image databases performing association rule mining over textual descriptions and automatically extracted features of the image content. Based on the knowledge mined, each query posed is rewritten in order to better meet the user expectations. We propose an extension of SQL aimed at exploring mining processes over complex data, generating association rules that extract semantic information from the textual description superimposed to the extracted features, thereafter using them to rewrite the queries. As a result, the system obtains results closer to the user expectation than it could using only the traditional, plain similarity query execution.


association rules; content-based retrieval; query rewriting; similarity queries; SQL extension; user expectation

Full Text:


An official publication of the Brazilian Computer Society Special Interest Group on Databases.