Research On Cross-domain App Requirements Acquisition Method Based On Text Analysis

With the development of the mobile Internet,smart phones have become popular,and the number of apps has increased exponentially.The competition in the App market has become extremely fierce,and the needs of users have become more complex.Developers need to continuously acquire innovative needs and develop more competitive products.As the initial stage of software requirements engineering,requirement acquisition mainly completes the requirement collection work,which is also a key factor in determining the success of the App.Traditional demand acquisition methods-such as face-to-face interviews,observation methods,and situational methods,have great difficulties in the face of uncertain demand in the App development cycle and difficulty in direct communication with users,which hinders them in the field of app demand acquisition The emerging demand acquisition methods,although text mining,sentiment analysis and domain analysis methods are used for the text and data of the App,the description text and comment data of the App are required to be acquired(such as SAFE [5],CHABADA),But often the vision of demand extraction is limited to the same product or the same field,making the innovative expansion of products very difficult.To this end,this article starts from the cross-domain information,breaks the limitation of the previous demand acquisition thinking,and proposes a cross-domain demand acquisition method based on grammar-directed translation(SDT).The entire requirement acquisition method can be classified into two parts,namely,the feature extraction method of App description text based on grammar-directed(SDT),and the domain requirement list recommendation model that integrates cross-domain information.In the first part,we propose an extensible feature extraction technology.This technology is based on the idea of grammatical-directed translation(SDT)in the compilation principle.By using Stanford Parser to construct a syntax analysis tree,we have carried out semantic analysis and analysis on massive App description texts.Rule extraction is to integrate the grammatical rules that contain the required features.Then,according to the integrated rules,the grammar tree of the new input text is hierarchically traversed.When the non-leaf node that meets the rules is traversed,the semantic action is embedded in the symbol In the stack,the grammatical actions in the symbol stack are finally executed in order to extract functional features from the App description text;the second part proposes a way to use the extracted feature information to provide crossdomain features from the supply domain to the receiving domain Recommend the list model.This part first uses the APP description text to train the Word2 vec language model and vectorizes the feature text.The vectorized result uses cosine similarity to quantify the relationship between features and generate a feature relationship network.After the network results filter the weak similarity features,the recommended range is narrowed through the idea of TF-IDF,and the remaining features are clustered based on the Euclidean distance DBSCAN,the clustering results are generated through a comprehensive evaluation function to generate a recommendation list,and the crossdomain is completed.The characteristics are recommended.Finally,through experiments,the effectiveness of each stage of the requirement acquisition method is verified.The accuracy,recall,and F-measure are selected to verify the effectiveness of the feature extraction method,and all of them have reached more than 80%.The final demand model can be reused.Different quantitative models are used for operability,operability and adaptability.The reusability is verified by Reuse Rank(RR)and the average value reaches 60%.The operability and adaptability are evaluated by Likert table,and the average score is Higher than 3 points,also reached a more ideal level.