Abstract

Having enough knowledge and keeping it up to date enables designers to execute the design assignment effectively and gives them a competitive advantage in the design profession. Knowledge elicitation or acquisition is a crucial component of system design, particularly for tasks requiring transdisciplinary or multidisciplinary cooperation. In system design, extracting domain-specific information is exceedingly tricky for designers. This thesis presents three works that attempt to bridge the gap between designers and domain expertise. First, a systematic literature review on data-driven demand elicitation is given using the Environment-based Design (EBD) approach. This review address two research objectives: (i) to investigate the present state of computer-aided requirement knowledge elicitation in the domains of engineering; (ii) to integrate EBD methodology into the conventional literature review framework by providing a well-structured research question generation methodology. The second study describes a data-driven interview transcript analysis strategy that employs EBD environment analysis, unsupervised machine learning, and a range of natural language processing (NLP) approaches to assist designers and qualitative researchers in extracting needs when domain expertise is lacking. The second research proposes a transfer-learning method-based qualitative text analysis framework that aids researchers in extracting valuable knowledge from interview data for healthcare promotion decision-making. The third work is an EBD-enabled design lexical knowledge acquisition framework that automatically constructs a semantic network -- RomNet from an extensive collection of abstracts from engineering publications. Applying RomNet can improve the design information retrieval quality and communication between each party involved in a design project.
To conclude, this thesis integrates artificial intelligence techniques, such as Natural Language Processing (NLP) methods, Machine Learning techniques, and rule-based systems to build a knowledge acquisition framework that supports manual, semi-automatic, and automatic extraction of design knowledge from different types of the textual data source.