In this thesis we propose the development of ontology-based services to facilitate the interoperability among heterogeneous agents that belong to a multi-agent system (MAS) dedicated to business-to-business (B2B) electronic commerce. These services will be integrated in an Electronic Institution to support the automatic negotiation required to the formation of Virtual Enterprises (VE). Our ontology-based services are provided through an Ontology-based Services Agent which is responsible for providing the agents within the MAS, representing customer and supplier enterprises, with suggestions of how to interpret and negotiate specific products, leading to meaningful conversations and making agreements possible. We address two types of interoperability problems: (i) problems caused by the use of distinct terminology; and (ii) problems generated by the lack of standardization. In this context, ontologies play an important role since they provide the vocabularies used by the agents. Whereas, in homogeneous MAS, the adoption of a common ontology guarantees the intelligibility of the communication between agents, in open MAS, where the agents are intrinsically heterogeneous and have different private ontologies, the interagent communication is affected by unknown and, thus, incomprehensible terminology, as well as, by the ad-hoc use of disparate unit systems to represent the attributes of products. The approach presented in this thesis aims at creating a methodology that matches two products represented in different ontologies through a lexical and semantic similarity evaluation. The lexical measure algorithm compares the characteristics (attributes), relations and descriptions of both products. The attributes are classified according to their data types and to the has-part relation. The concepts (names) of both products are compared using the Leacock-Chodrow WordNet-based semantic similarity measure algorithm. To handle the question of the lack of standardization, a set of units conversion Web services are provided.