Session Outline
The talk provides an overview of how to effectively evaluate recommender systems, considering various properties such as the ability to discover new items or explore diverse items rapidly. The different types of experiments that can be used to assess these properties, such as offline, online and user studies are discussed. The challenges of conducting a comprehensive evaluation of a recommender system and the exciting future directions in this area are also addressed.
Key Takeaways:
- The importance of comprehensive evaluation in understanding the full picture of the success of a recommender system.
- The various properties that can impact the success of a recommendation system and the appropriate experiment types for evaluating these properties.
- The challenges and pitfalls of evaluating recommender systems and the future directions for work in this field.
