In this work, we present a method to build a data conceptual schema from information verbalized in
natural language. This method has three components : a process, based on models, and realized by
softwares. We use two conceptual models : the Entity-Relationship model and the « elementary
semantic model », which allows us to represent texts expressed in natural language. The process is
divided into steps guiding and controlling the construction of conceptual schemas from texts. It is based
on a linguistic approach obtained by formalizing the linguistic mechanisms used by analysts to translate
the Universe of Discourse into concepts. This linguistic approach is realized by softwares aimed to
mimic the process used by an analyst to produce a conceptual schema form texts.