ABSTRACT
Web pages) based on their user profile [1] which can be explicit (e.g., user ratings) or implicit (e.g., their browsing or purchase history). One of the most successful forms of recommendation is collaborative filtering (CF) [2], which recommends to the current user, items that are of interest to users who are similar to this user, where similarity is based on a correlation-like affinity between the user’s ratings or purchasing histories. For this reason, some people think of recommendation technology as a new paradigm of search, since interesting items find the user instead of the user
actually searching for them.Asa result, recommendersystemsarebecominga requisite stapleonmanye-commerceWebsites,andareboundtoconnecttogetherthefuturesocial Web.Good recommendation systems thriveontheavailabilityofdata (e.g.,Web sitesor products viewed or purchased) to form profiles of the users. These profiles are, in a way, starting to play the role of the user’s own digital persona in the Internet marketplace. Yet, despite their success and their broad potential, to this day, unfortunately, a loyal consumer/user of a Web business still cannot own their profile or their persona. Neither can they move it with them freely, as they move from one business or one context to another. However, this profile forms extremely precious information, both to the business and to the users, and it can benefit personalization and combat information overload in a variety of different domains for the following reasons:
1. There are likely correlations between a user’s tastes in books, movies, and many other products or content items that are not sold on the same Web site, including food, wine, clothing, sports, arts, content-like news and blogs, as well as music, videos, etc. Hence there is a need for single-profile integration across multiple Web sites.
