User-centered Evaluation of Recommender Systems
DEI PhD Student
DEI - 3B Room
November 14th, 2011
Recommender Systems play an increasingly important role in on line applications characterized by a very large amount of data – e.g., multimedia catalogs of music, products, news, images, or movies. Their goal is to filter information and to recommend to users only the items that are supposed to be the most appealing to them. The quality of a RS is typically defined in terms of different attributes, the principal ones being relevance and serendipity. Relevance refers to the capability of providing items that fit the user’s preferences. Serendipity indicates the capability of suggesting new, fortuitous, or unexpected items. Most existing works in RS evaluation operazionalize quality in terms of statistical properties - error metrics and accuracy metrics – that do not involve users, i.e., quality is evaluated "algorithmically". The goals of this work is to explore whether algorithmic measures of RS quality are in accordance with user-based measure, and to investigate at which degree such statistical metrics are reliable predictors for the quality actually perceived by users.
Web, multimedia and databases