CPref-SQL: uma linguagem de consulta com suporte a preferências condicionais: teoria e implementação

Abstract

Many important applications like e-commerce and recommendation systems require the use of ecient techniques for customizing and manipulating user preferences. In the database eld, the researches are focused on the development of query languages able to express and lter preferences. In fact, database systems that integrate user preferences have been a good solution for personalization and high quality in queries answers. In this dissertation we propose the implementation of the CPref-SQL query language, an SQL extension that supports conditional preferences. In this language, the preferences are specied by a set of rules expressed in terms of a logical formalism. To this end, we generalize the CPref-SQL preference model, making it more expressive and develop the algorithms G-BNL and GRank-BNL, which implement the new preference operators Select-Best and SelectK-Best, respectively. These operators are able to evaluate top-k queries with preferences, i.e., queries that return the K most preferred tuples according to a preference hierarchy. Finally, we implemented the language in the core of the PostgreSQL query processor. An extensive set of experiments demonstrates the eciency of the methods used in language development. Given the implementation, we also present case studies that illustrate the application of the CPref-SQL language in a context of integration between similarity queries and user preferences.