Abstract: The term “Semantic Web” indicates a set of programs that utilize semantic tags for the information retrieval within texts diffused on the Web. Insertion of semantic tags is enabled by ontologies, enormous collections of specific definitions about contents, organized in classificatory schemas. The implementation of “ontologies” allows different domains to interact, so that it is possible to make various searches that correlate different types of texts. Semantics involved in semantic web today refers mainly to semantic definitions and related ontologies that essentially concern proper and/or common names. Secondly, semantics is intended as a formal logical apparatus in order to hierarchically organize the conceptual system of Class, sub-Class and properties. This paper is a contribution to the debate on SW in a new perspective based on the possibility to elaborate Semantic Predicates starting from concrete lexical forms and their syntactic structures in which they are collocated within informative and communicative contexts. Such Predicates would be automatically generated within texts by a lexical-grammatical “intelligent agent”. Texts tagged in this way could be queried on their contents in natural language.