Abstract

Requirements elicitation and modeling are critical for the success of product development not only in software engineering but also in other engineering fields. Collecting the right requirements at each stage and transforming them into conceptual models are essential in delivering a successful product. In most cases, original requirements are represented by natural language in engineering. However, a key challenge faced by industries is to transform existing loosely structured legacy requirements document into the structured representations. This transformation process is extremely time-consuming and prone-to-error. Some efforts in research have been made to develop automatic or semi-automatic processes to bridge natural language and formal representation. Motivated by both the strong industrial need to automatically formalize natural language based requirements (NLR) and the research breakthrough in product requirements modeling, this present thesis proposes a new approach to transforming product requirements from their unrestricted natural language representation to structured conceptual models by using Recursive Object Model (ROM).
The proposed approach includes the following three main aspects: 1) developing criteria for the completeness and necessity of design requirements corresponding to certain design stage, 2) developing a dynamic requirements elicitation approach to refine requirements, and 3) developing algorithms for transforming design requirements from natural language to conceptual models, such as Use Case Model by Universal Modeling Language (UML) and Function-Behavior-State (FBS) model. This presented research involves Natural Language Processing (NLP) techniques, in conjunction with question asking (QA) strategy and conceptual modeling algorithms. The significant tasks include defining the scope of the right requirements, automatically question asking to elicit requirements, formulating the transformation of requirements text into conceptual models, generating the rules for the conceptual modeling, developing algorithms based on the transformation rules, and finally automating the requirements modeling process through software prototypes.
The research foundation of this thesis is the Environment Based Design (EBD) methodology which is derived from axiomatic theory of design modeling (ATDM). To bridge the gap between unrestricted natural language and formal conceptual models, an intermediate representation, ROM, is the core for representing the semantics of design requirements throughout the requirements evolution process.