Linguagens de consulta para banco de dados com suporte a preferências condicionais

Abstract

Nowadays, the treatment of preferences has become an important task in several kinds of applications like e-commerce and personalized search engines. This arised a lot of research in the Articial Intelligence eld as well as in the Database eld. In the Articial Intelligence eld, several formalisms for preference specication and reasoning have been created. In the Database eld, the research has been focused on extending SQL to support preference specication. In this dissertation, we use preference formalisms originated in the AI eld in order to specify the query languages CPref-SQL and TPref-SQL. The CPref-SQL query language consists of an extension of the standard SQL language with the operators Best-E, Best-N and Best-D which select the tuples of a relation taking into account a set of conditional preferences specied by the user. We propose algorithms for implementing these operators and implement a prototype for a fragment of the CPref-SQL language. The TPref-SQL query language is based on a relational model where data are stored in sequence relations. In the context of this sequence relational model, we introduce a Relational Algebra with special constructors allowing to operate on sequence relations. Among these constructors we have the three operators BestSeq-E, BestSeq-N and BestSeq-D (the counterparts of the CPref-SQL operators Best-E, Best-N and Best-D) allowing to select the best sequences in a sequence relation, according to a set of temporal conditional preferences. We propose algorithms for implementing these operators and implement a prototype for a fragment of the TPref-SQL language.