Session: CIE-25-01 - Graduate Student Poster Symposium

Paper Number: 97938

97938 - Natural Language Processing to Model Based Systems Engineering in the Architecture Analysis and Design Language (Aadl) 

Model based engineering has enabled automated analytical reasoning early in the design phase. As a result, discrepancies and design flaws can be identified early on in the development process. However, a gap exists between the natural language-based specifications and their actual implementation. This is because formal method-based tools rely on mathematical concepts and computing theories that necessitate specialized knowledge, limiting the applicability of model-based engineering. Natural language is the most widely used method to represent specifications. So, it is intuitive to utilize natural language-based representation to generate system and formal annotations such that it will enable automated architectural analysis with much wider acceptance. The objective of this research is to develop the above-mentioned approach that integrates representation of the specifications in a subset of English language which can then be used to generate system architecture in AADL along with the generation of functional specifications. We can validate our approach with use cases from the aerospace and electromechanical domains.  The main motivation of this work is to automate the process of generating a system architecture model from natural language specifications text. Natural language is the initial and the most essential
step in the representation of the technical specification of a system or software. It is also the most widely used method for communicating the specifications. But natural language-based representation is prone to interpretability issues, ambiguity, interoperability, and it does not capture inconsistencies in the design of systems, unless a formalized grammar is developed that enables automated analytical reasoning. We propose to develop an approach that addresses the gap between the specifications written in English and the design developed using formal methods and systems engineering principles. The impact, however, is bigger than just automating the entire process. The secondary motivation of this research is to develop a tool to aid system engineers to generate architecture models through natural language text.

Considering the gaps stated previously, this research aims to answer the following research questions.
RQ1. Can we automate the process of generating a system architecture model from natural language
specifications?
RQ2. How can we update the NLP field of research to include a corpus for function modeling
specification terms for better POS tagging?
RQ3. How does this approach compare to other architecture generation and analysis approaches?

The result of this research will include an end-usable software tool, which will allow software developers and system engineers in general, to accelerate the start of designing the system architecture based on requirements text given by the user. Our approach also shows a more generic framework which has the potential to have a broader impact on modeling and analysis of systems, as it is not specific to a particular domain. Ultimately, developing this approach in a commercial platform such as AADL would pave the way to integrate it as a plugin for an Annex in AADL.

This research will develop a framework that will automate the process of creating an AADL model representing the structural, functional and behavior aspects of a system described in natural language text. In addition, we will validate this framework by testing it on the system requirements given by a system design engineer. Furthermore, this approach will help designers visualize some functional aspects of a system in AADL which is not currently supported.

Presenting Author: Parth Ganeriwala Florida Institue of Technology

Presenting Author Biography: Parth Ganeriwala earned his B.E. in Computer Science from Birla Institute of Technology and Sciences, Pilani in December 2021. He is now pursuing his M.S. in Computer Science at Florida Institute of Technology. His research<br/>interests include deep learning, transfer learning, formal methods and artificial intelligence. He is a graduate research assistant at the ASSIST (Assured Safety, Security and Intent with Systematic Tactics) Research Lab working towards the development of assurance frameworks for mission-critical, safety-critical, and security-critical systems.

Authors:

Parth Ganeriwala Florida Institue of Technology