Abstract: The present invention provides an improved recommender system. In some embodiments, the recommender system automatically builds a user profile for a user based solely on the action or actions the user takes with respect to certain documents and the contents of the documents. That is, the profile is based on implicit ratings. In addition to maintaining a user profile for each user, each document in a set of documents will have its own document profile, which is also referred to as a document vector. Advantageously, in some embodiments, the user profile will consist of the same attributes as the document vectors. This advantageous feature enables the recommender system to compare users and documents in order to generate recommendations.

Abstract: A method of providing content (130) associated with a weight-value, the content (130) previously provided to a current computer associated with a current user (116) that is represented by a first node (106) in a social network (101). The method comprises the steps of: i) enabling the current computer (216) to display the content (130), in dependence of the weight-value, ii) obtaining an input of the current user (116), iii) updating the weight-value of the content (130), in dependence of the input of the current user (116), iv) determining a receiving computer associated with a second node (107) in the social network (101), and v) providing the content (130) to the receiving computer. Corresponding computers, computer program and computer readable medium are also described.

Abstract: The invention provides an improved recommender system that includes a client device or service provider server, a trusted function handler module and a recommender module. The recommender system functions to protect the privacy of user rating information maintained by the node (i.e., client device/server) by having the node transform the user rating information using a specific function selected by the function handler and then provide the transformed user rating information to the recommender module. In this way, privacy of the user rating information is maintained because the original user rating information will be unknown to a recommender module.

Abstract: A method is provided for determining a correlation between a reference user and another user on the basis of two sets of ratings, where each rating is associated with a respective user. In response to a trigger, user ratings associated with the reference user and user ratings associated with the other user are collected, and all co-rated items of these two sets are correlated in the basis of an adjusted cosine correlation function which is weighted by a first and a second weighting function. The correlation is then stored and may be repeated for a plurality of users. The stored correlations may be used e.g. for ranking purposes.

Abstract: In one embodiment, a method of generating annotation tags (28) for a digital image (22) includes maintaining a library (16) of human-meaningful words or phrases organized as category entries (72) according to a number of defined image description categories (70), and receiving context metadata (20) associated with the capture of a given digital image (22). The method further includes selecting particular category entries (72-1, 72-2) as vocabulary metadata (24) for the digital image (22) by mapping the context metadata (20) into the library (16), and generating annotation tags (28) for the digital image (22) by logically combining the vocabulary metadata (24) according to a defined set of deductive logic rules (30) that are predicated on the defined image description categories (70). In another embodiment, a processing apparatus (12), such as a digital processor (18, 26) and supporting memory (14), etc., is configured to carry out the above method, or to carry out variations of the above method.

Abstract: A method of estimating a default rating of a rated dataset is provided, where the dataset comprises at least one series of ratings associated with at least one user and each series comprise ratings associated with at least two items. For a reference user and an item for which a rated value is missing the item's average rating, ir, the reference users average rating, Ru, and the datasets average rating, dr, is collected. A Poisson distribution of the reference users rating is then generated on the basis of the reference users average rating. A random Poisson rating, ur, is calculated on the basis of the Poisson distribution, and a default rating, r, is estimated by weighting the random Poisson rating on the basis of the items average rating, the users average rating and the datasets average rating.

Abstract: The invention provides an improved recommender system that includes a client device or service provider server, a trusted function handler module and a recommender module. The recommender system functions to protect the privacy of user rating information maintained by the node (i.e., client device/server) by having the node transform the user rating information using a specific function selected by the function handler and then provide the transformed user rating information to the recommender module. In this way, privacy of the user rating information is maintained because the original user rating information will be unknown to a recommender module.

Abstract: Methods, devices, and computer-readable media described herein may provide a recommender system that may increase the serendipity associated with a recommendation. The recommender system omits obvious co-occurred items, rare items, and limits a number of co-occurred items associated with an item table. Local and global weighting values may be calculated to derive a co-occurrence weight. The co-occurrence weight may be compared to maximum and minimum threshold co-occurrence values to omit obvious and rare co-occurred items.

Abstract: In one embodiment, a method of generating annotation tags (28) for a digital image (22) includes maintaining a library (16) of human-meaningful words or phrases organized as category entries (72) according to a number of defined image description categories (70), and receiving context metadata (20) associated with the capture of a given digital image (22). The method further includes selecting particular category entries (72-1, 72-2) as vocabulary metadata (24) for the digital image (22) by mapping the context metadata (20) into the library (16), and generating annotation tags (28) for the digital image (22) by logically combining the vocabulary metadata (24) according to a defined set of deductive logic rules (30) that are predicated on the defined image description categories (70). In another embodiment, a processing apparatus (12), such as a digital processor (18, 26) and supporting memory (14), etc., is configured to carry out the above method, or to carry out variations of the above method.

Abstract: A method of providing content (130) associated with a weight-value, the content (130) previously provided to a current computer associated with a current user (116) that is represented by a first node (106) in a social network (101). The method comprises the steps of: i) enabling the current computer (216) to display the content (130), in dependence of the weight-value, ii) obtaining an input of the current user (116), iii) updating the weight-value of the content (130), in dependence of the input of the current user (116), iv) determining a receiving computer associated with a second node (107) in the social network (101), and v) providing the content (130) to the receiving computer. Corresponding computers, computer program and computer readable medium are also described.

Abstract: A method of estimating a default rating of a rated dataset is provided, where the dataset comprises at least one series of ratings associated with at least one user and each series comprise ratings associated with at least two items. For a reference user and an item for which a rated value is missing the item's average rating, ir, the reference users average rating, Ru, and the datasets average rating, dr, is collected. A Poisson distribution of the reference users rating is then generated on the basis of the reference users average rating. A random Poisson rating, ur, is calculated on the basis of the Poisson distribution, and a default rating, r, is estimated by weighting the random Poisson rating on the basis of the items average rating, the users average rating and the datasets average rating.

Abstract: A method is provided for determining a correlation between a reference user and another user on the basis of two sets of ratings, where each rating is associated with a respective user. In response to a trigger,user ratings associated with the reference user and user ratings associated with the other user are collected, and all co-rated items of these two sets are correlated in the basis of an adjusted cosine correlation function which is weighted by a first and a second weighting function. The correlation is then stored and may be repeated for a plurality of users. The stored correlations may be used e.g. for ranking purposes.

Abstract: The present invention provides an improved recommender system. In some embodiments, the recommender system automatically builds a user profile for a user based solely on the action or actions the user takes with respect to certain documents and the contents of the documents. That is, the profile is based on implicit ratings. In addition to maintaining a user profile for each user, each document in a set of documents will have its own document profile, which is also referred to as a document vector. Advantageously, in some embodiments, the user profile will consist of the same attributes as the document vectors. This advantageous feature enables the recommender system to compare users and documents in order to generate recommendations.