RANKING NEWS FEED ITEMS USING PERSONALIZED ON-LINE ESTIMATES OF PROBABILITY OF ENGAGEMENT
An on-line social network system includes a ranker to processes an inventory of news feed updates for a member and select more relevant updates for presentation to the member. The ranker is trained using training data that includes personalized engagement probability for an update. The personalized engagement probability values are calculated in real time and for a particular update with respect to member features that appear in member profiles maintained by the on-line social network system.
1 . A computer-implemented method comprising:
collecting in real time engagement signals with respect to a focus update in an on-line social network system, an engagement signal represents a positive engagement with the update or an absence of engagement with the update by a member from the focus members, the focus members are those members in the on-line social network system to whom the update has been presented, the focus members represented by respective member profiles each including one or more features from a set of features, the focus update is an information item for presentation to one or more members represented by respective member profiles in the on-line social network system;
calculating, using at least one processor, in real time, for the focus update and the set of features, respective probabilities of positive engagement with the focus update by a member represented by a profile that includes a feature from the set of features, based on the collected engagement signals; and
including the calculated respective probabilities as training data for training a final pass ranker, the final pass ranker to generate a rank for each item in an inventory of updates based on features associated with a member profile in the on-line social network system.
2 . The method of claim 1 , comprising selecting one or more items from the inventory for including them in a news feed of the member represented by the member profile, based on the respective ranks generated by the final pass ranker for each item in the inventory of updates.
3 . The method of claim 2 , comprising constructing a news feed web page that includes the one or more items from the inventory.
4 . The method of claim 3 , comprising causing presentation of the news feed web page on a display device of the member.
5 . The method of claim 1 , comprising calculating, for the focus update and one or more features from the set of features, a probability of positive engagement with the focus update by a member represented by a profile that includes the one or more features.
6 . The method of claim 5 , comprising utilizing the calculated probability of positive engagement with the focus update by a member represented by a profile that includes the one or more features as training data for training the final pass ranker.
7 . The method of claim 5 , wherein the calculating, for the focus update and the one or more features from the set of features, the probability of positive engagement with the focus update by a member represented by a profile that includes the one or more features comprises utilizing a logistic regression model.
8 . The method of claim 5 , wherein the calculating, for the focus update and the one or more features from the set of features, the probability of positive engagement with the focus update by a member represented by a profile that includes the one or more features comprises utilizing a Naive Bayes model.
9 . The method of claim 1 , wherein the focus update is a job posting or a news article.
10 . The method of claim 1 , wherein a feature in the set of features represents an item selected from a group comprising an industry, a company, a geographic location, and a skill.
11 . A computer-implemented system comprising:
an engagement signals collector, implemented using at least one processor, to collect engagement signals with respect to a focus update in an on-line social network system, an engagement signal represents a positive engagement with the update or an absence of engagement with the update by a member from the focus members, the focus members are those members in the on-line social network system to whom the update has been presented, the focus members represented by respective member profiles each including one or more features from a set of features, the focus update is an information item for presentation to one or more members represented by respective member profiles in the on-line social network system;
a personalized engagement probability calculator, implemented using at least one processor, to calculate, in real time, for the focus update and the set of features, respective probabilities of positive engagement with the focus update by a member represented by a profile that includes a feature from the set of features, based on the collected engagement signals; and
a training module, implemented using at least one processor, to include the calculated respective probabilities as training data for training a final pass ranker, the final pass ranker to generate a rank for each item in an inventory of updates based on features associated with a member profile in the on-line social network system.
12 . The system of claim 11 , comprising a selector, implemented using at least one processor, to select one or more items from the inventory for including them in a news teed of the member represented by the member profile, based on the respective ranks generated by the final pass ranker for each item in the inventory of updates.
13 . The system of claim 12 , comprising a web page generator, implemented using at least one processor, to construct a news feed web page that includes the one or more items from the inventory.
14 . The system of claim 13 , comprising a presentation module, implemented using at least one processor, to cause presentation of the news feed web page on a display device of the member.
15 . The system of claim 11 , wherein the personalized engagement probability calculator is to calculate, for the focus update and one or more features from the set of features, a probability of positive engagement with the focus update by a member represented by a profile that includes the one or more features.
16 . The system of claim 15 , wherein the training module is to include the calculated probability of positive engagement with the focus update by a member represented by a profile that includes the one or more features as training data for training the final pass ranker.
17 . The system of claim 15 , wherein the personalized engagement probability calculator is to calculate, for the focus update and the one or more features from the set of features, the probability of positive engagement with the focus update by a member represented by a profile that includes the one or more features comprises utilizing a logistic regression model.
18 . The system of claim 15 , wherein the personalized engagement probability calculator is to calculate, for the focus update and the one or more features from the set of features, the probability of positive engagement with the focus update by a member represented by a profile that includes the one or more features comprises utilizing a Naive Bayes model.
19 . The system of claim 11 , wherein the focus update is a job posting or a news article.
20 . A machine-readable non-transitory storage medium having instruction data executable by a machine to cause the machine to perform operations comprising:
collecting in real time engagement signals with respect to a focus update in an on-line social network system, an engagement signal represents a positive engagement with the update or an absence of engagement with the update by a member from the focus members, the focus members are those members in the on-line social network system to whom the update has been presented, the focus members represented by respective member profiles each including one or more features from a set of features, the focus update is an information item for presentation to one or more members represented by respective member profiles in the on-line social network system;
calculating, in real time, for the focus update and the set of features, respective probabilities of positive engagement with the focus update by a member represented by a profile that includes a feature from the set of features, based on the collected engagement signals; and
including the calculated respective probabilities as training data for training a final pass ranker, the final pass ranker to generate a rank for each item in an inventory of updates based on features associated with a member profile in the on-line social network system.