The number of available Android applications continues to increase, leading to increasing pressure on developers to release popular applications. As the default distribution channel for Android apps, Google Play (GP) contains over 3.8 million Android apps. However, due to a lack of discoverability, users may be unaware of the available app features. Imagine the frustrating user experience when a user is playing an Android game app without any description or how-to-play guide. Therefore, for GP Android app developers, it is crucial to make app features clear and discoverable. Especially, improving app discoverability can help vision-impaired people save time in manually exploring app functionalities.

Within this project, we will first conduct a large-scale discoverability analysis in real-world Android apps. Based on the empirical study findings, we will design algorithms and tools to improve the discoverability of mobile apps. On the one hand, we hope to design a tool that can automatically generate high-quality app descriptions based on other information such as screenshots, reviews, videos, or even apk files by leveraging deep learning. On the other hand, we will propose an approach to identify the runtime discoverability issues and recommend potential suggestions.