Abstract: One or more server devices may obtain a stored resource identifier associated with a resource identifier collection. The one or more server devices may further access a document index that stores information about canonical resource identifiers, where the information relates a particular canonical resource identifier to one or more other resource identifiers; determine whether the canonical resource identifier differs from the stored resource identifier; and replace the stored resource identifier with the canonical resource identifier, when the canonical resource identifier differs from the stored resource identifier.

Abstract: Methods, program products, and systems for classifying feeds are described. A system acquires user created labels and user subscribed data feeds from a group of users. The system generates a topic-feed association rule by generating a topic-feed mapping that contains topic-feed pairs, calculating a weight for each topic-feed pair, and selecting topic-feed pairs whose weight reaches a threshold. The data feeds are classified into topics according to the topic-feed association rules. The system obtains user categories from a group of users. Using a probability model, the system matches the user subscribed data feeds with the user categories. The system further generates a topic classification, associating topics with user categories. Using the topic classification, topic-feed association rules, and a user's specific data, the system predicts a user's feeds of interest and recommends data feeds to the user accordingly.