Research On Intelligent Requirement Acquisition And Modeling Method For E-collaboration System

In the background of world economic integration and the domestic economic cycle era,E-collaboration(EC)systems,as a tool for creating virtual work environments,have become the main way for people to break through the boundaries of time and space to achieve information interaction and resource sharing,and are widely used in education,business,transportation,medical care,and so on.However,since the user scale of EC system is expanding and the targets of related stakeholders are becoming more and more complex,the development of EC system faces the following challenges: 1)how to quickly and effectively obtain the use cases of EC system requirements;2)how to meet the targets of different related stakeholders and accurately and completely model the EC system;3)how to efficiently and timely explore new user requirements to update and iterate the system.In order to effectively respond to the above challenges,this paper proposes an intelligent requirement acquisition and modeling method for EC systems based on deep learning,and the research includes three stages: UML(Unified Modeling Language)use case graph matching,EC system requirement modeling and system newly added requirement acquisition,and the specific work is carried out as follows.(1)The key to efficiently acquire EC system requirement use cases is to match the most similar EC use case graph models,and in the similarity matching of UML use case graph models,most of the previous studies have matched the similarity of models from a single perspective and at high cost.To address the appeal problem this paper proposes a similarity matching method for use case graphs oriented to requirements reuse,the main steps of which are as follows: 1)Use XML file parser to parse the XML model file of the use case graph to obtain the elements of the use case graph(i.e.,Actor,Use Case,and Relationship);2)Use the obtained data to perform similarity matching in the use case graph repository,and perform similarity calculation from both structural and semantic dimensions.The structure aspect first converts the XML model into a directed graph,and then uses the Sim GNN neural network to calculate the similarity score of the corresponding graph structure.The semantic aspect extracts the descriptions of key elements in the use case graph and converts them into keyword sets,and obtains the similarity score by combining cosine similarity with TF-IDF;3)Finally,the semantics and structure are given certain weights to calculate the final similarity score,and after getting the most similar use case graph model,its use case description can be obtained.(2)In the aspect of EC system requirement modeling,traditional system design specifications usually involve the static structure of the system,which ignores the most basic characteristics of EC systems —— 4Cs characteristics(Coordination,Communication,Collaboration,and Connection),and the 4Cs characteristics of EC systems lead to the fact that designing their requirement models is not an easy task.In response to the appeal problem,this paper proposes an EC system requirement modeling method based on 4Cs characteristics,whose main steps are as follows: 1)Get the most similar use case graph model by UML use case graph similarity matching method and obtain its use case description;2)Use NLP(Natural Language Processing)techniques to extract relevant information,such as noun phrases and verb phrases,from EC system use case description texts;3)Mapping the extracted information to the modeling elements,and using the Process model,Agent model and Goal model to model the coordination,communication,connection and collaboration characteristics of the EC system pairs respectively,in order to support the EC system design more completely.(3)In the aspect of EC system new requirements acquisition,how to automate the mining of the requirements of the stakeholders for the collaboration and real-time system functions is the focus and difficulty of the research.To solve the appeal problem this paper proposes a deep learning based EC system new requirement acquisition method,whose main steps are as follows: 1)Summarize the requirement categories proposed by previous researches,combine the characteristics of EC system to make a more detailed division of functional and non-functional requirements,better cover the requirement categories involved in the development process of EC system,so that EC system developers can more accurately get the direction of users’ requirements;2)Crawl user review information from the APP application market,clean,annotate and simplify the data to form a collaboration system requirement review classification dataset;3)Use Text Rank ranking algorithm and Text CNN model to classify them,and finally by building a network of co-occurrence relationships between words to feed the users most concerned in a certain classification to the developers.To evaluate the proposed method in this paper,experiments are conducted on a dataset constructed by domain experts and on review data collected in the Apple App Store,respectively.The experimental results show that the method in this paper have better performance to a certain extent.