| 摘要 |
A recommendation network is described. The recommendation network may include, but is not limited to, users or entities ("recommendation sources,") that explicitly or implicitly recommend, rate or refer items and users or entities that receive recommendations or referrals ("recommendation receivers,"). Users who make recommendations can create multiple recommendation sources, using different recommendation sources for different purposes, by assigning different recommendations to different recommendation sources. The recommendation network allows recommendation receiver to assign relative trust ratings to recommendation sources, the relative trust rating representing the recommendation receiver's confidence that the recommendation source, compared to other recommendation sources, makes recommendations the recommendation receiver considers valuable. The relative trust ratings can be ordinal or cardinal values that can be used by the network algorithm to filter and rank items for the recommendation receiver. The network may rank items for the recommendation receiver based on the number of recommendation source referencing the item and the relative trust rating the recommendation receiver has assigned to those recommendation sources. By ranking items for each recommendation receiver according to the a ranking algorithm that uses the relative trust ratings the recommendation receiver has assigned to those recommendation source, the network allows the recommendation receiver to receive recommendations for items in the form of a ranked list, the items being ranked by the degree they are recommended by the recommendation receiver's trusted recommendation sources. By allowing the recommendation receiver to assign relative trust ratings to different recommendation source, and by ranking and displaying recommendations by the degree to which they are recommended to the recommendation receiver, the network allows users to subscribe to, or receive, recommendations from a limitless number of recommendation sources without being overwhelmed. The network prioritizes the recommendations for an recommendation receiver by the parameters the recommendation receiver establishes, imposing order on a potentially-limitless number of pushed recommendations. According to an embodiment of the invention, a recommendation receiver can be a recommendation source, and can create recommendation sources, for themselves or other recommendation receivers. The recommendation receiver may explicitly or implicitly recommend items as a recommendation source, and can also recommend other recommendation sources, or re-label other recommendation sources, or rate multiple recommendation sources and combine them, to create new recommendation sources. Besides combining multiple recommendation sources to create a new recommendation source, recommendation receivers can create new recommendation sources by restricting the new recommendation source to include only those recommendations, from one or more existing recommendation sources, that share or avoid particular characteristics. The new recommendation sources may in turn be used by the recommendation receiver, or other recommendation receivers. Other recommendation receivers can assign relative trust values to the new recommendation sources. Therefore, by interpreting or reinterpreting received recommendation sources to create new recommendation sources, then by recommending those new recommendation sources, the recommendation receiver provides more information, which helps other recommendation receivers to identify valuable recommendations and information about items.
|
