M. Ehresmann, J. Beyer, S. Fasoulas

Requirement documents are the basis for software developments in space engineering projects. They contain specifications on properties of the result or product. Precision of the stated requirements drives the probability that the final result is viable for the commissioner. In the work reality, space projects - already on software level - consider thousands of requirements. The majority of requirements are stated in natural language. During most projects individuals transfer requirements to individuals with subsequent manual and to a degree subjective interpretation. Due to project complexity catching misinterpretations early on is challenging, possibly leading to an increase in cost and development duration. Manual transfer and transformation to software tools is a significant work load. Processing requirements readable by machines is therefore a potential for optimizing the design process, reducing cost, effort, and error rate. ExANT (Extraction of requirements from natural language texts) is a project that evaluates the feasibility of using artifical intelligence natural language processing for processing software specifications to enhance the digitization of the software development process. The consortium consists of space software development experts from Gerlich System and Software Engineering (GSSE), software engineers from OHB System AG and OHB Digital Services, as well as the Institute of Space Systems University of Stuttgart (IRS) to combine academic and industry knowledge and expertise. The scope of this project considers additionally to being a feasibility study, potential for cost and effort reduction or at least the possibility to enhance requirement quality in order to increase clarity and precision. For computerized natural langauge processing open-source tools as well as commercially available tools are evaluated. Prior to training the respective neural network, the preparation of data in natural language texts and its annotation is required. This is necessary to be able to later correctly derive contents, structure and logical contexts of given requirements. Preliminary considerations for natural language processing of software specifications and the respective annotation process with the help of IBM Watson are given in this paper.

Deutscher Luft- und Raumfahrtkongress 2022, Dresden

Deutsche Gesellschaft für Luft- und Raumfahrt - Lilienthal-Oberth e.V., Bonn, 2022

Conference Paper

englisch

21,0 x 29,7 cm, 7 Seiten

urn:nbn:de:101:1-2022111813111664035966

10.25967/570134

Artificial Intelligence, Natural Language Processing, Requirements Engineering

Download - Bitte beachten Sie die Nutzungsbedingungen dieses Dokuments: CC BY 4.0OPEN ACCESS

Ehresmann, M.; Beyer, J.; Fasoulas, S. (2022): Artificial Intelligence for Cost Reduction in Space Engineering: Extraction of Software-Requirements from Natural Language Requirement Documents. Deutsche Gesellschaft für Luft- und Raumfahrt - Lilienthal-Oberth e.V.. (Text). https://doi.org/10.25967/570134. urn:nbn:de:101:1-2022111813111664035966.

18.11.2022