From Natural Language Requirements to a Conceptual Model

In literature it is described in great detail how class diagrams and ER diagrams or UML class diagrams are derived from natural language sentences. It is normally assumed, that there is a direct correspondence between natural language elements (e.g., words) and conceptual model elements. We do not strictly follow this assumption because of the complexity of natural language with its ambiguities and ellipsis. Hence in this paper a stepwise generation of a conceptual model out of natural language requirements sentences is proposed. According to the ideas of MDA we assume that automatic transformation steps from the source model (in our case natural language) to the target conceptual model (e.g., UML class diagram) make sense. In addition to that we suggest that the designer should play an important part during transformation. It is furthermore proposed to introduce an interlingua which helps to detect defects and provides traceability between sentences and the model elements. Index Terms – natural language processing, interlingua, conceptual modeling, defect detection