Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
241938 | Advanced Engineering Informatics | 2014 | 9 Pages |
Abstract
In this article we report on validated research for the construction of design constraints by automated interpretation of natural language input. We show how our approach of dynamic reconfigurations of parsed syntax trees using a number of production rules is used to formalize and transform natural language constructs into computable constraints that are applied to concrete building information models. The calibration and validation of the proposed algorithms and production rules is based on two test data sets: The verbatim text of Dutch building code regulations for dormer extensions on existing roofs and constraints formulated ad hoc by design students based upon a series of example designs. We show how a prototypical implementation of our approach can be used to interpret 44% of the test data without human interference and how the remaining sentences can be interpreted with minimal additional effort or further development.
Related Topics
Physical Sciences and Engineering
Computer Science
Artificial Intelligence
Authors
R.A. Niemeijer, B. de Vries, J. Beetz,