Frappe

Icon512

With almost one million different apps in Google Play, it has become very difficult for users to discover great and relevant apps. Particularly apps that are not ranked on the top lists or niche apps that are very useful in specific situations. Frappé helps you find these apps.

Frappé is a context-aware app discovery tool that will recommend the right apps for the right moment. We use it mainly to test our developed algorithms and to compare it with the state of the art. At the moment it runs algorithm described in this paper (pdf). Frappé is continuously learning in which context users are using their favorite apps. Moreover, it analyses the user bahaviour within the application and use this as a feedback loop. By leveraging all the knowledge Frappé can recommend the right apps for the right moment.

Frappé is transparent and will explain why an app was recommended to you. Maybe your co-workers have an app installed and use it in the office? Are you at home on weekend and many people like you are playing a trending game you didn't hear yet? Frappé will recommend it!

Key features:

  • Most advanced application discovery tool available.
  • Recommendations will change to fit the context around you.
  • Frappé keeps learning becomes better with time and keeps track of the newest and trending apps.
  • Customizable as you can select which categories you like most, if you want to receive notications etc.
  • Helps you to discover great paid apps that became free for a limited time.
  • Personalized results that are different form the top-pop results of Google Play.
  • Categories are also context-aware to help you find great and useful apps.
  • Interfaces are designed for usability with a clean and beautiful style.
  • Easy to use and forever free, try it now and let us know your opinion.

 

The code

The code for frappe can be found in these repositories:

  1. https://github.com/grafos-ml/frappe contains the serving engine
  2. https://github.com/grafos-ml/test.fm contains the models

Data Set:

The anonymized frappe data set that can be downloaded from HERE. It contains 96202 records by 957 users for 4082 apps. Please cite the following paper if you use the data:

@article{frappe15,
  author    = {Linas Baltrunas and
               Karen Church and
               Alexandros Karatzoglou and
               Nuria Oliver},
  title     = {Frappe: Understanding the Usage and Perception of Mobile App Recommendations
               In-The-Wild},
  journal   = {CoRR},
  volume    = {abs/1505.03014},
  year      = {2015},
  url       = {http://arxiv.org/abs/1505.03014},
  timestamp = {Mon, 01 Jun 2015 14:13:54 +0200},
  biburl    = {http://dblp.uni-trier.de/rec/bib/journals/corr/BaltrunasCKO15},
  bibsource = {dblp computer science bibliography, http://dblp.org}
}

We hope that the release of this data set will further push the state of the art in context-aware recommender systems.

 

Research:

We use frappe as a testbed for our research on recommender systems.

  • Y. Shi, A. Karatzoglou, L. Baltrunas, M. A. Larson, and A. Hanjalic, (2014) "CARS2: Learning Context-aware Representations for Context-aware Recommendations". In Proc. CIKM '14, 291-300, ACM.
  • L. Baltrunas, K. Church, A. Karatzoglou, N. Oliver, (2015) Frappe: Understanding the Usage and Perception of Mobile App Recommendations In-The-Wild. arXiv:1505.03014