We present Mobipedia, an integrated knowledge base with information about 1 million mobile applications (apps) such as their category, meta-data (author, reviews, rating, release date), permissions and libraries used, and similar apps. The goal of Mobipedia is to integrate unstructured and semi-structured data about mobile apps from publicly available data sources and publish it as Linked Data using RDF. We describe the extraction process for facts, access mechanisms to the knowledge base, and an overview of applications facilitated by Mobipedia.
The number of mobile applications (also called apps) available for various platforms has seen an exponential growth in the last few years (for example, the Google Play Store achieved the 1 million apps milestone in 2013). This has resulted in smart phones replacing other devices as de facto medium for online browsing, social networking, and other activities. Today's users have a wide array of choices while nding apps for entertainment, utility, or education.
However, this huge number of apps has also made the choice of an appropriate app di cult. There are many parameters to be taken into account when selecting an app such as technical (such as the version of the operating system supported, the hardware required, or the installation size), user experience (such as ratings and comments), and privacy concerns (such as the information that the app would access or the third-party libraries used). As a matter of fact, different studies have been performed on app stores and some of them have publicly released their datasets and results. But these projects are mostly isolated from one another and scattered across the Internet. In addition, the use of di erent methods to release the datasets (e.g., websites, dumps, or databases) and different formats (from unstructured to semi-structured data) has made accessing them di cult.
Through Mobipedia4, we envision an evolving knowledge base (KB) containing information related to mobile apps. Mobipedia integrates information from various sources such as o cial websites, and research projects. In this paper we introduce the current status of Mobipedia describing the ontology created to model knowledge about mobile apps, the di erent sources that have already been integrated, the access mechanisms o ered, and an overview of the applications which can be developed using Mobipedia. We believe that having an online knowledge base integrating information about mobile apps would accelerate the research in various domains related to mobile apps for e.g., mobile privacy, and app search. 2
Mobipedia Dataset To create the dataset we utilized the information available for apps on the Google Play Store. Each app's metadata includes information such as their category, images, version, installation size, developer, comments and permissions used. We created classes, and data, and object properties to model all this information. We have also included additional information (other than what is available on the Play Store) about mobile apps from open datasets such as PlayDrone and PrivacyGrade mentioned below (e.g., libraries used by each app and developer metadata). Figure 1 shows an excerpt of the ontology including the most important classes and the object properties that relate them5. 4 http://mobipedia.link 5 The gure has been generated using the Gra oo speci cation http://www.
essepuntato.it/graffoo/
Information Extraction. To populate the ontology with instances we extracted
facts from two research projects, PlayDrone [
From these sources we extracted information about more than 1M apps and added them as RDF triples in the Mobipedia KB. Each of these entities are described in the dataset by a URI of the following form where entity corresponds to an app, developer, permission, rating and so on: http://mobipedia.link/ontology/entity .
Accessing Mobipedia. Similarly to DBpedia [
Linking Mobipedia with other knowledge bases. We have linked Mobipedia with DBpedia. Speci cally, with instances of the DBpedia categories Mobile software and Android (operating system) software. 6 https://github.com/yasserg/crawler4j 7 http://owlapi.sourceforge.net 8 https://github.com/primalpop/MobipediaProject 9 http://systems.cs.columbia.edu/projects/playdrone 10 http://privacygrade.org 11 http://developer.android.com/reference/android/Manifest.permission.html In Mobipedia, we have focused on creating a single point of access for Android app related data, which can be easily accessed through access mechanisms mentioned earlier. We believe that a Linked Data cloud of Mobile apps would make it easier to develop applications which utilizes app data and have outlined some of them below.
{ Semantic search portal for mobile apps: Enable users to nd relevant apps based on their semantic search criterion. For instance, sports games with parental control; to-do list with location reminders, or ashlight with least number of required permissions. { Detection of ad targeting: With the information of ad libraries being used by apps, permissions requested, and developer metadata that Mobipedia stores it could be possible to draw inferences about which app developers are \going rogue" with respect to targeting users for ads. { Linking application user experiences: The information of user app experiences such as app ratings, reviews, blog articles, forums and so on while using an application is fragmented across various sources. Using Mobipedia, this information can be linked to apps itself, which could be leveraged to build smarter app recommendation systems.
Mobipedia has to adapt to the dynamic nature of mobile app stores with new apps and new version of existing apps being released almost daily. Therefore, Mobipedia will be a continuously evolving knowledge base by incorporating these new information. We also intend to link the entities in Mobipedia to other open and popular datasets like Freebase.
Acknowledgments. This research work has been supported by RADICLE project CNS-1059436, CNS-1212943, CNS-1118127 and CNS-1450768, CICYT project TIN201346238-C4-4-R and DGA FSE, U.S. National Science Foundation awards 0910838 and 1228198.