This is a tracking task for our project to improvement the current article recommendation system in the Android app.
**Goal:** This project aims to increase user engagement on the Android app. The engagement should be increase by providing a better article recommendation system, which assists users in accessing relevant articles.
**Approach:** We want to develop a framework that allows the evaluation of recommender systems (RS) with respect to our goal. The framework requires the integration of additional RS into the MediaWiki/CirrusSearch API, the use of these recommendation in the Android app and measurement of performance metrics. A fixed success criterion should determine whether a RS out-performs the current system.
At first, the link-based recommendations from [[ https://github.com/wikimedia/citolytics| Citolytics ]] should be tested.
**Sub-tasks:**
# T149682 Define success criteria
# T142831 Review the current automatic recommendation mechanism (Lucene/MoreLikeThis)
# T148833 Use Android Analytics framework for collecting click through-data
# T143197 Integrate additional RS (Citolytics) to CirrusSearch
# Evaluate how the [Click stream based evaluation](https://github.com/wikimedia/citolytics) measures change
**Issues**:
- citolytics currently works on article level. Figure out if there is clickstream data on section level
**Time frame** (deprecated)
- 20.10.2016 Finish implementation (ok)
- 05.11.2016 Start code review process
- 24.12.2016 finish data collection
- 24.01.2016 finish evaluation
We will update this ticket step by step.