Bugine
The figure shows the overall workflow of our issue recommendation system, Bugine. Our approach first builds a database of GitHub issues obtained from open-source Android apps and pre-processes these issues to extract their metadata. For each app in our database, we extract its app description file for future comparison. Given an app under test Appquery, Bugine extracts its UI components to obtain its app description file. Then, we use the similarity between the app description file for Appquery and the app description files for all the apps in our database to search for apps that are similar to Appquery. This similarity is given as input to our ranking function, which prioritizes the GitHub issues in our database. Finally, Bugine output a ranked list of relevant GitHub issues for Appquery.