Abstract : To find their way in "territories" of complex activities with huge, specific and moving glossaries and shared meanings, human actors need maps. It is important to help them to use, to organize and to modify again and again the "topic map" linking topics describing their domain, their activity and their position as actors. Besides, there often exist several points of view; the topic map has to take this plurality into account: we name it a "Semiotic Ontology" and we present the Hypertopic model and the "Knowledge-Based Marletplace model (KBM) to represent this collectively constructed knowledge. Nowadays many researches are focusing on a "Semantic Web" to provide better services. In this context of improving the Web standards, they should provide such "Hypertopic Maps" to improve content management and information retrieval. According to us, the more efficient and useful Semantic Web, understood as a shared meaning artifact using Web standards, is the one created within and by communities: it is a "Socio-Semantic Web", by which the community could build a dynamic and collective meaning.