The "Recommender-Systems in Action Meetup" aims at learning about recommender systems in the real-world, i.e. recommender systems that are being used by real users. We are less interested in talking about supposedly super-performing novel approaches that were only evaluated with offline evaluations or small user studies in artificial lab environments.
We want to have talks that help to answer the following questions:
- Recommender System Evaluation: What are the objectives your company is trying to achieve with a recommender system? How do you measure the effectiveness of a recommender system? What do you do, when different metrics report contradicting results? What are the challenges in evaluating recommender systems, and how do you tackle those challenges?
- Recommender System Infrastructure: On what kind of machines is your recommender-system running (cloud, dedicated servers, ...)? How many people work on your recommender system? How much does/did it cost to develop and maintain the system? Which recommendation frameworks (Mahout; Apache Lucene; ...) and which other technologies do you use? How long does it take to e.g. index new documents?
- Recommender System Innovation: What did you do, to improve your recommender system's effectiveness? Which recommendation approaches did you use, and which ones were best? How did you improve existing recommendation approaches? How did you deal with e.g. the cold start problem?
- The little things: How many recommendations do you display to users? How often? Through which media (website, email, ...)? How do you motivate your users to use the recommender system and e.g. generate ratings?
- Recommender System Demos: Have you implemented a recommender system? Then show it to us in a short live demonstration.
- Recommender Systems Failures: What were the biggest failures you experienced when developing and running a recommender system?
- Recommender Systems Careers: Do you have vacant jobs or internships, or are you looking for an exciting project for your Master or PhD thesis? Then present your vacancy or yourself.
The typical structure of a meetup will be as follows:
Up to 4 speakers present their company or university and their job offers, internship opportunities, PhD positions, or Masters projects. In addition, up to 4 speakers who are looking for collaborators and industry partners may present themselves and e.g. their PhD projects.
- Presentations (10-20 minutes each, not more than 1h in total)
3-5 speakers talk about their latest insights into recommender systems, addressing, for instance, one of the above-mentioned questions.
- Demos (5 minutes each)
Up to three presenters demonstrate their real-world recommender system in action.