Improvement in the Translation Process from Natural Language to System Dynamics Models

Abstract

In order to solve social science problems or make models for business forecasts, descriptive information in natural language is as important as measured numerical information. While numerical information is widely used in various stages, most descriptive information is employed only to describe the background or the very primitive stage of modelling. However, descriptive information has rich contents. One uses it to organise one’s ideas and to communicate with other people. The modelling method which can employ this rich content information is System Dynamics. It has an interface to express such descriptive information, stock flow diagrams. Moreover, these diagrams are also used when numerical simulations are conducted. However, matching descriptive information and diagrams has depended on model builders’ skill and mental models. This paper shows matching rules between descriptive information and stock flow diagrams as improvement in an existing method.