Semantic information exchange is currently facing the heterogeneity and consistency across languages context in a smart city. Thus, this paper highlights the challenges that smart city is confronting in terms of semantic document exchange and proposes a novel Machine Natural Language Mediation (MNLM) framework which provides a sentence-based machine natural language (MNL) as a kind of mediation language, where each sentence as a compound concept is a set of atomic concepts to be compatible with all languages achieving a global semantic transformation. The MNL enables the sentence computer-readable and understandable through unique iid and eiid without semantic ambiguity, and simultaneously MNLM will eliminate the semantic inconsistency and improve the accurate meaning interpretation across conversational contexts. The overall aim of the framework is to offer a better solution to communicate and exchange information or document in a smart city.