Abstract:
This thesis proposes an approach for automating the content negotiation of services as an alternative to price-oriented negotiation approaches. In content-oriented negotiation, the description of services rather than their price are negotiated. Both consumers and producers share an ontology about the service of interest. The shared ontology contains the features for a given service and captures the relations between them as well as providing the representation of semantics. Through repetitive interactions, the provider learns consumers’ needs accurately and can make better targeted counter offers. In order to learn the consumer preferences, several approaches are explored such as the extension of Version Space, the use of decision trees in an incremental way and the combination of learning with a variety of semantic similarity metrics. The main contributions of this thesis are the extension of Version Space with the capability of learning the disjunctive concepts and a new semantic similarity metric using the taxonomies. The proposed architecture for the negotiation combines important ideas from incremental learning techniques with the expressive representation of ontologies. In addition to the theory of the automated negotiation architecture, the details of the implemented system are given. A variety of learning algorithms and similarity metrics are tested. The test results show the constructive effect of considering the semantic closeness among services.