José Tomás Cadenas

Fuzzy domains with adaptable semantics in an object-relational DBMS

In this paper, we focus on the management of the users context in the field of Fuzzy Database applications. Concretely, we propose an approach that uses fuzzy techniques to represent vague concepts with an adaptable semantics in order to meet users preferences. We enhance previous work on the design and implementation of fuzzy databases, allowing a flexible use of different sets of linguistic labels for the same domain, which are transparently chosen according to users context. This way, applications that operate on fuzzy databases get an additional piece of semantic power that makes an automatic customization in relation to users preferences possible. This initial capability is proposed within a general architecture that aims at the management of the user context from a more general and interdisciplinary point of view. We complete our contribution with the description of a first proof of concept implementation of our proposal using the Oracle Object-Relational DBMS.

About fuzzy query processing

Traditional Relational Database Management System (RDBMS) query languages suffer from rigidity. There are some proposals for solving this problem using fuzzy sets, such as SQLf. The trouble with fuzzy querying languages is that they have been proposed to be implemented with loosely or middle coupling architecture, adding undesired overhead. This paper deals with the theoretical and pragmatic foundations for providing SQLf with tight coupling. We present an original experimental study that evidences performance of tight coupling approach for fuzzy querying.

About the Performance of Fuzzy Querying Systems

Traditional database systems suffer of rigidity. Use of fuzzy sets has been proposed for solving it. Nevertheless, there is certain resistance to adopt this, due the presumption that it adds undesired costs that worsen the performance and scalability of software systems. RDBMS are rather complex by themselves. Extensions for providing higher facilities, with a permissible performance and good scalability would be appreciated. In this paper, we achieve a formal statistics study of fuzzy querying performance. We considered two querying systems: SQLfi (loose coupling) and PostgreSQLf (tight coupling). Observed times for the later are very reasonable. It shows that it is possible to build high performance fuzzy querying systems.