| The software interface,also known as the User Interface(UI),serves as the interface between users and applications(App).It is one of the essential factors affecting users’ evaluation of applications.However,the evaluation of UI quality mainly depends on users’ subjective feelings,making it difficult to judge whether the UI meets users’ expectations through simple tests.Therefore,developers should pay attention to users’ feedback on applications during the maintenance process and actively respond to their feedback.App user reviews in application stores provide developers with a way to obtain users’ suggestions on UI.There is a wealth of valuable information hidden in App user reviews,providing new opportunities for collecting user opinions and improving the UI in a timely manner.The purpose of mining UI-related information in App user reviews is to provide valuable information for developers to maintain apps from the perspective of user experience,so as to improve user experience.This thesis uses App user review data to identify UI topics,then analyzes the identified UI topics.The specific main contents are as follows:(1)A method for expanding the user interface dictionary is proposed.Existing research has identified some UI-related keywords from a large number of App user reviews.To obtain more UI-related keywords,first,collect the review data used in the experiment.Then,use the word2 vec word embedding model to obtain the word vectors of each word in the reviews,and calculate the cosine similarity between the word vectors in the review dataset and the UI-related keyword vectors obtained from previous research.If the value is greater than the set threshold and is approved through group discussion,the word is considered a similar word.Finally,compare the keywords obtained by previous researchers and the number and quality of UI-related reviews obtained after adding similar words and filtering keywords.Experimental results show that the derived similar words are effective.(2)A UI topic recognition method based on App user reviews is proposed.First,analyze the number and rating of UI-related reviews to roughly understand users’ opinions on the UI.Then,preprocess the noise in the App user reviews.Next,compare and analyze the topic-perplexity curve and the schematic diagram of topics generated by py LDAvis to determine the appropriate number of topics,and then use the topic model to perform topic modeling on UI-related reviews.Experimental results show that the number of topics obtained by py LDAvis is more intuitive and has more practical significance.(3)The association between UI topics and interface design principles is studied.First,label UI-related reviews as positive or negative according to the rules,and then use sentiment analysis to predict the sentimental tendency of UI topics.Then,map the top ten interface design principles into three categories of UI-related issues.Finally,associate UI topics with UIrelated issues and draw conclusions based on the sentimental tendency of topics.Experimental results show that users care about the appearance,usability,and fault-tolerance of the UI.Among them,usability receives the most attention,and frequent updates to the UI can lead to negative user reviews. |
