Abstract: A highlight learning technique is provided to detect and identify highlights in sports videos. A set of event models are calculated from low-level frame information of the sports videos to identify recurring events within the videos. The event models are used to characterize videos by detecting events within the videos and using the detected events to generate an event vector. The event vector is used to train a classifier to identify the videos as highlight or non-highlight.

Abstract: A highlight learning technique is provided to detect and identify highlights in sports videos. A set of event models are calculated from low-level frame information of the sports videos to identify recurring events within the videos. The event models are used to characterize videos by detecting events within the videos and using the detected events to generate an event vector. The event vector is used to train a classifier to identify the videos as highlight or non-highlight.

Abstract: A highlight learning technique is provided to detect and identify highlights in sports videos. A set of event models are calculated from low-level frame information of the sports videos to identify recurring events within the videos. The event models are used to characterize videos by detecting events within the videos and using the detected events to generate an event vector. The event vector is used to train a classifier to identify the videos as highlight or non-highlight.

Abstract: Methods, systems, and apparatus include computer programs encoded on a computer-readable storage medium, including a method for providing content. One or more content items are identified that are eligible to be presented in response to received requests in accordance with one or more selection criteria. One of the content items includes one or more parameters that can be varied. A range is identified over which a parameter of the one or more parameters is variable, and a subset of the range is identified, wherein the subset is less than the entire range. A request for content is received that is satisfied by the content item. The content item is prepared, including randomly selecting a value for the parameter that is within the subset of the range. At least one performance aspect associated with the prepared one content item is monitored.

Abstract: A highlight learning technique is provided to detect and identify highlights in sports videos. A set of event models are calculated from low-level frame information of the sports videos to identify recurring events within the videos. The event models are used to characterize videos by detecting events within the videos and using the detected events to generate an event vector. The event vector is used to train a classifier to identify the videos as highlight or non-highlight.

Abstract: A highlight learning technique is provided to detect and identify highlights in sports videos. A set of event models are calculated from low-level frame information of the sports videos to identify recurring events within the videos. The event models are used to characterize videos by detecting events within the videos and using the detected events to generate an event vector. The event vector is used to train a classifier to identify the videos as highlight or non-highlight.

Abstract: A system and method for determining related content to serve based on connectivity is provided. The system includes a connectivity analyzer to analyze an aspect of the connectivity of the detected access to the content sharing service; a vector aggregator to create an access vector based on the analyzed aspect of the connectivity; and a related content retriever to retrieve related content based on the access vector.

Abstract: Methods and systems for suggesting one or more semantic tags for a media clip are disclosed. In one aspect, a media clip provided by a user is identified, and a first set of semantic tags is generated for the media clip based on a feature vector associated with the media clip. The first set of semantic tags is then provided to a classifier that is trained based on user selection of semantic tags. Further, a second set of semantic tags is obtained from the classifier and is suggested to the user for the media clip.

Abstract: Disclosed are methods for automatically generating commands to transform a video sequence based on information regarding speaking participants derived from the audio and video signals. The audio stream is analyzed to detect individual speakers and the video is optionally analyzed to detect lip movement to determine a probability that a detected participant is speaking. Commands are then generated to transform the video stream consistent with the identified speaker.

Abstract: A video classification score boosting method boosts classification scores for videos for increased accuracy. A target video is classified with a classifier, producing a classification score. Related video scores are determined using the classifier for sets of videos related to the target video. The sets of related videos may include co-browsed videos, co-commented videos, co-queried videos, and co-uploaded videos. The related video scores may be the mean or median classification score for the classified sets of related videos. Weighting coefficients associated with the classifier are retrieved and applied to the classification score and the related video scores. The weighting coefficients may be determined for the classifier by classifying sets of pre-classified videos with the classifier and determining the weighting coefficients which, when applied to the classification scores of the pre-classified videos, improves the accuracy of the classification scores.