Spotify Ab

Erik Bernhardsson

Kevin Faaborg

Yinghao Li

A method of selecting content items based on user interactions with provided content is performed at a server system. A subset of content items selected from a plurality of available content items is provided to a first user of the service. An amount of time that the subset of content items is provided to the first user is determined. Content item selection criteria for a second user of the service is adjusted based on the amount of time that the subset of content items is provided to the first user. In response to receiving a request for a content item from the second user, a content item is selected items for the second user from the plurality of available content according to the adjusted content item selection criteria. The selected content item is provided to the second user.

Systems and methods of classifying content items

A method of classifying content items utilizes user-generated playlists. A content item is included in playlists from a plurality of distinct users. A server system receives, from respective users of the plurality of distinct users, respective playlist titles for the playlists. In accordance with a determination that the respective playlist titles from a threshold number of the plurality of distinct users specify a first attribute, the server system assigns the first attribute to the content item. The server system receives a request, from a first user via an application that is executing on a client device that is distinct from the server system, for content having the first attribute. In response to receiving the request, the server system selects the content item according to the first attribute and delivers the selected content item for playback to the first user.

A method of classifying content items utilizes user-generated playlists. A content item is included in respective playlists from a plurality of respective distinct users. The method receives respective user-generated information corresponding to the content item from each of the respective distinct users. For some users, the respective user-generated information is the respective playlist title. For other users, the user-generated information is the text of social network posting that identifies a respective playlist. In each case, the respective user-generated information specifies a first content item attribute that characterizes the content item. Accordingly, the method assigns the first content item attribute to the content item. Subsequently, a request is received from a first user for a content item having the first content item attribute. In response, the method selects the content item according to the first content item attribute and delivers the first content item to the first user.

A method of selecting content items is performed by an electronic device having one or more processors and memory. The memory stores one or more programs for execution by the one or more processors. The method includes providing a first content item to a first user. The first content item is selected from a plurality of available content items, such as audio tracks or videos. The first user provides feedback relating to the first content item, and the feedback is used to adjust content item selection criteria for a second user distinct from the first user. The method includes receiving a request for a content item from the second user and selecting a content item from the plurality of available content items for the second user according to the adjusted content item selection criteria. The selected content item is then provided to the second user.

Systems and methods of selecting content items

A two-dimensional matrix of data points represents occurrences of users selecting content items, such as audio tracks. A server computes n-dimensional latent vectors for each user and for each item. The server iteratively optimizes the user vectors and item vectors based on the data points. Each iteration includes a first phase in which the item vectors are held constant, and a second phase in which the user vectors are held constant. In the first phase, the server computes first phase parameters based on data points, the user vectors, and the item vectors, and updates the user vectors. In the second phase, the server similarly computes second phase parameters for the item vectors and updates the item vectors. The server receives a request from a user for an item recommendation, and selects an item vector based on proximity in n-dimensional space. The server then recommends the selected item to the user.

Growing community of inventors

New York, NY, United States of America

Erik Bernhardsson