Hercules Dalianis, Anders Hedman and Paul Johannesson
Abstract
The STEP/EXPRESS standard is one of the most complex ever developed. Understanding its workings is not an easy task for new users. Designers and other persons involved in the modelling process need to validate STEP/ EXPRESS specifications. They need, therefore, a tool generating natural language descriptions from STEP/EXPRESS specifications. This paper investigates the differences between the STEP domain and other domains with respect to natural language generation architectures. In order to do this, a questionnaire was handed out to personnel at Volvo Data and DSV-KTH-SU asking how to paraphrase parts of a PDM (Product Data Model) into Natural Language text. The texts were then analysed using aggregation and discourse techniques. The findings were implemented in a tool that automatically generates natural language descriptions from STEP/EXPRESS specifications derived from Application Protocol 214 Core Data for Automotive Design Process.