Task ontologies are used as a communication bridge to handle semantic inconsistencies between the same attributes used at different sites. The main achievement of this paper is its presentation of a rough-sets query answering system based on distributed data mining.

The paper is divided into five sections. After an introduction, section 2 recalls the notion of a distributed information system and knowledge base for a client site, formed from rules extracted at a remote site. The notion of local queries and examples of local semantics are then presented.

In section 3, the authors introduce the notion of a distributed autonomous knowledge system (DAKS), and present problems related to the construction of its query answering system (QAS). The authors also demonstrate how global queries can be processed if two sites involved in a query resolution use different granularity levels for the same attribute.

In section 4, the notion of a reduct is recalled. The authors also demonstrate how the partially ordered set of semantics can be used to improve the query answering process in DAKS.

The authors present several interesting examples. They conclude that it may be considered rather unrealistic to choose one local semantics for incomplete databases as a standard. They propose that in such cases, a partially ordered set of local semantics (&OHgr;, ⪯), or its equivalent structure, be a part of task ontologies, to provide this new metadata information as needed to solve inconsistency problems.

The paper includes 18 references, and will be of interest to readers involved in ontology-based systems.