Design and Implementation of a Google Action enabled Smart Agent System for Mobile App Review Monitoring based on Sentiment Analysis Techniques

Antonio Fernández. (2017). Design and Implementation of a Google Action enabled Smart Agent System for Mobile App Review Monitoring based on Sentiment Analysis Techniques. Trabajo Fin de Titulación (PFC). ETSI Telecomunicación, Universidad Politécnica de Madrid.

Abstract:
Nowadays, app development for smart phone ecosystem is quite fast and easier than ever. A freelance developer can have an idea, implement it by himself and publish the app in the market with a low barrier of entry. As a mobile product developer, it is essential to place your apps in the top of the store rankings, maintaining your product by implementing new features and minor bug fixes. However, iOS and Android market platforms only provide customer feedback with an average rating of 1 up to 5, and sometimes including a short review with the user experience. In most cases, this information is not enough to identify why your app has a bad acceptance, and consequently a bad positioning in the store. This lack of data provided by marketplaces about how user feels using an app means a problem for developers. Mining the reviews of a certain application, and performing a sentiment or emotion analysis, we can obtain an implicit more detailed information about the product, identifying how the user feels, and evaluate his accordance with the product. Even comparing same user reviews in similar apps, we can extract some information of what do my users like about other related apps?, and make recommendations based on this data. Furthermore, bug and features classification it’s also a good implementation, so developer teams composed by low number of programmers can redirect the feedback obtained from the store previously tagged. To carry out these solutions we have defined several objectives. First of all, develop a smart agent for Android devices where user can perform sentiments and emotions analysis for different Play Store apps. The smart agent interacts with the user through voice or text sentences, which will be converted into specific actions using a Natural Language Understanding system. Moreover, the system allows us to automate Play Store review response for applications developed by the user, depending on the classification result obtained. Finally, offering a service oriented to developers, where they can manage and analyze mobile application market status from their Android terminal.