Abstract: An online system is configured to recommend a live video to a target user by presenting the live video in the target user's newsfeed or other location. The online system receives a plurality of live videos for streaming to its users. To recommend a live video, the online system dynamically identifies a topic of each of the plurality of live videos during the streaming. The online system further selects a group of candidate live videos, each of which has an identified topic matching an interest of the target user. The online system ranks the candidates based on a probability that the target user would positively interact with each candidate live video if the candidate live video is presented to the target user. Then the online system provides a candidate above a threshold level in the ranking for display to the target user, e.g., in a newsfeed of the target user.

Abstract: An online system is configured to recommend a live video to a target user by presenting the live video in the target user's newsfeed or other location. The online system receives a plurality of live videos for streaming to its users. To recommend a live video, the online system dynamically identifies a topic of each of the plurality of live videos during the streaming. The online system further selects a group of candidate live videos, each of which has an identified topic matching an interest of the target user. The online system ranks the candidates based on a probability that the target user would positively interact with each candidate live video if the candidate live video is presented to the target user. Then the online system provides a candidate above a threshold level in the ranking for display to the target user, e.g., in a newsfeed of the target user.

Abstract: An online system that recommends content to a user obtains a content item and receives limited interaction with the content item when it is obtained. The online system also maintains embeddings for various users based on interactions by the users with content presented to the users. Viewing users to whom the content item is presented and who performed one or more interactions with the content item are identified. Based on embeddings for the viewing users, the online system generates a content embedding for the content item. When the online system identifies a candidate user, the online system determines a similarity between the content embedding and an embedding of the viewing user and determines whether to recommend the content item to the candidate user based on the comparison.

Abstract: An online system generates personalized video heat map for a target user, which visually indicates segments of a video likely to be of interest to the target user. The online system divides the video into the segments and identifies actions performed by users other than the target user on each of the segments. The online system determines embedding vectors describing each segment as represented by the identified actions performed on the segment and also determines an embedding vector describing the target user. Based on those embedding vectors, a personalized score for the segment is determined using a trained model. The online system uses the personalized score for each segment of the video to generate the personalized heat map. The personalized heat map can be provided to the target user along with the video.

Abstract: An online system is configured to recommend a live video to a target user by presenting the live video in the target user's newsfeed or other location. The online system receives a plurality of live videos for streaming to its users. To recommend a live video, the online system dynamically identifies a topic of each of the plurality of live videos during the streaming. The online system further selects a group of candidate live videos, each of which has an identified topic matching an interest of the target user. The online system ranks the candidates based on a probability that the target user would positively interact with each candidate live video if the candidate live video is presented to the target user. Then the online system provides a candidate above a threshold level in the ranking for display to the target user, e.g., in a newsfeed of the target user.

Abstract: An online system that recommends content to a user obtains a content item and receives limited interaction with the content item when it is obtained. The online system also maintains embeddings for various users based on interactions by the users with content presented to the users. Viewing users to whom the content item is presented and who performed one or more interactions with the content item are identified. Based on embeddings for the viewing users, the online system generates a content embedding for the content item. When the online system identifies a candidate user, the online system determines a similarity between the content embedding and an embedding of the viewing user and determines whether to recommend the content item to the candidate user based on the comparison.

Abstract: An online system generates personalized video heat map for a target user, which visually indicates segments of a video likely to be of interest to the target user. The online system divides the video into the segments and identifies actions performed by users other than the target user on each of the segments. The online system determines embedding vectors describing each segment as represented by the identified actions performed on the segment and also determines an embedding vector describing the target user. Based on those embedding vectors, a personalized score for the segment is determined using a trained model. The online system uses the personalized score for each segment of the video to generate the personalized heat map. The personalized heat map can be provided to the target user along with the video.