Research And Application On A Semi-Automatic Requirements Modeling Method Based On Domain Knowledge

Over the last decade,researchers have extensively researched iStar,an effective approach for requirements modeling.However,until now,iStar has not been widely used in practice.Researchers generally agree that the scalability issues in large-scale requirements modeling hinder the practical application of the iStar framework.There have been many research works that aim to address the framework scalability issues.However,they obtained little effect.The reason for such results is that these works did not systematically and comprehensively analyze the needs before proposing solutions and ignored the voice of practitioners.Many modeling tools are available in the iStar framework,divided into two categories:graphical modeling tools and textual modeling tools.However,few researchers have ventured into the abstract modeling methods buried under the modeling tools.A modeling tool is the external expression of a modeling method,but its essence is still the modeling method.The issue of tool scalability needs to solve from the root,which is the modeling method.In terms of layout algorithms for iStar models,in recent years,some research works have proposed layout algorithms for iStar models based on the force-directed layout method.These works are admittedly quite innovative theoretically,but none of them have systematically analyzed the visualization and layout requirements of the iStar model in practical scenarios before proposing,and ultimately their contributions in practical applications are rather limited.Currently,researchers are generally aware of the scalability issues of the iStar framework,but few research works have proposed solutions for the target users.Towards alleviating the scalability issues and promoting the practical application of the iStar framework,this dissertation improves the iStar framework in terms of modeling methods and model layout.These improvements aim to pragmatically reduce the learning threshold and usage cost of the iStar framework,alleviate the scalability issues,and improve the efficiency of modeling and analysis.This dissertation first systematically designed and implemented a controlled experiment involving 38 participants,and the results revealed the strengths and weaknesses of the existing iStar modeling methods.Based on the results,the dissertation systematically designed and conducted a case study involving 8 participants and obtained the requirements from modeling beginners and modeling experts to improve the iStar modeling method.Based on the domain knowledge from the first two empirical studies,this dissertation systematically designed and proposed a semi-automatic iStar modeling method to meet the requirements of the modelers.Finally,this dissertation systematically elicited iStar model visualization requirements,proposed a two-level force-directed layout algorithm to meet these requirements,and supported the proposed modeling method.The details of the research in this dissertation are as follows.Firstly,this dissertation systematically designed and conducted a controlled experiment using modeling tools to compare two typical iStar modeling methods(graphical vs.textual)and to mine the characteristics of the modeling methods from the advantages and disadvantages of the modeling tools.The experiment involved 38 participants and obtained the strengths and weaknesses of the two modeling methods.The results show that the graphical modeling procedure is more intuitive but suffers from scalability issues.And the textual modeling procedure is more logical but too unintuitive to use when there is only a textual model without a visual one.In addition,the experiment carried out a questionnaire to obtain the preferences for the two methods from participants.Secondly,this dissertation then systematically designed and conducted a case study for the target users of the iStar framework to elicit requirements from five beginners to iStar modeling and three domain experts with ten years of iStar modeling experience.Based on the domain knowledge gained from the case study,this dissertation proposed a user-friendly semi-automatic iStar requirements modeling method,combining the advantages of existing iStar modeling methods obtained in the aforementioned controlled experiment.The proposed method optimizes the shortcomings of existing methods and applies automation techniques in the process reasonably to improve modeling efficiency.The method supports the batch addition of model elements with a semi-structured model syntax and automatic visualization of the added model elements.Natural language processing techniques are also used in the proposed method to identify model elements in requirement documents automatically.The method contains two typical semi-automatic modeling processes.One is for the early-phase requirements analysis,starting from scratch,and the other is a semiautomatic iStar modeling process using requirements documents as input.The proposed method satisfied the requirements from modelers.Finally,this dissertation systematically collected and analyzed iStar model visualization requirements and proposed a two-level force-directed layout algorithm using a partitioning strategy according to the visualization requirements of the model.This dissertation systematically designed and conducted a case study involving 24 experts in the requirements engineering field to evaluate the understandability of the automatic layout results from the proposed algorithm,which was eventually generally accepted by the experts.The layout algorithm was also integrated into a leading opensource iStar modeling tool,evaluating the fulfillment of visualization requirements of the algorithm.It proved that the proposed layout algorithm fulfilled all the visualization requirements.After the integration,the proposed layout algorithm serves a wide range of researchers in the requirements engineering field.