Topic and time based media affinity estimation
An affinity server estimates an affinity between two different time based media events (e.g., TV, radio, social media content stream), between a time based media event and a specific topic, or between two different topics, where the affinity score represents an intersection between the populations of social media users who have authored social media content items regarding the two different events and/or topics. The affinity score represents an estimation of the real world affinity between the real world population of people who have an interest in both time based media events, both topics, or in a time based media event and a topic. One possible threshold for including a social media user in a population may be based on a confidence score that indicates the confidence that one or more social media content items authored by the social media user are relevant to the topic or event in question.
1. A computer-executed method, comprising:
accessing a plurality of content items authored by users of a social networking system;
accessing a plurality of annotated events of time based broadcast media, the events having been aired during segments of time, the annotated events comprising metadata instances that have semantic information about the events;
generating mappings between the content items and the annotated events, and between the content items and a plurality of topics;
determining a first affinity score between a first population of the users who authored content items mapped to a first one of the events or a first one of the topics and a second population of the users who authored content items mapped to a second one of the events or a second one of the topics;
determining a second affinity score between a third population of the users who authored content items mapped to a third one of the events or a third one of the topics and a fourth population of the users who authored content items mapped to a fourth one of the events or a fourth one of the topics;
determining a third affinity score between the second population and the fourth population; and
inferring a fourth affinity score between the first population and third population based on the first, second, and third affinity scores.
2. The computer-executed method of claim 1 , wherein generating the mappings comprises
performing a feature extraction process on pairs including one of the content items and one of the events or one of the topics, the feature extraction generating content features that identify co-occurring textual information between the content item and the metadata instances of the annotated event or one or more keywords associated with the topic of the pair; and
generating scores for the content features indicating whether the content items refers to the annotated events or topics using a feature specific sub-function for the content feature.
3. The computer-executed method of claim 1 , further comprising:
accessing programming information regarding the events of time based broadcast media; and
annotating the events using an alignment function to estimate the likelihood that the events are described by the metadata instances.
4. The computer-executed method of claim 1 , wherein determining one of the affinity scores comprises:
determining an intersecting population of the users between one of the populations mapped to one of the events or topics, and another one of the populations mapped to the one of the events or topics;
determining the affinity score based on the intersecting population.
5. The computer-executed method of claim 4 , wherein determining one of the affinity scores comprises calculating a weighted sum over the users in the intersecting population.
6. The computer-executed method of claim 5 ,
wherein a weight is assigned to each user in the intersecting population, and
wherein each of the weights is determined based on at least one from the group consisting of:
a number of times the user has authored one of the content items associated with one of the mappings, and
a probability that the user will author a content item relevant to either the event or topic associated with one of the populations forming the intersecting population.
7. The computer-executed method of claim 5 , wherein determining the affinity score comprises normalizing the weighted sum by an average probability of users to author content items regarding the topics or events associated with intersecting population.
8. The computer-executed method of claim 5 , wherein determining the affinity score comprises normalizing the affinity score based on a number of users in the populations forming the intersecting population.
9. The computer-executed method of claim 8 , wherein determining the affinity score comprises normalizing the affinity score based on at least one from the group consisting of:
a cardinality of the union between the users of the populations forming the intersecting population and
a product of the cardinality between the populations forming the intersecting population.
10. A non-transitory computer-readable storage medium comprising instructions that when executed cause a processor to:
access a plurality of content items authored by users of a social networking system;
access a plurality of annotated events of time based broadcast media, the events having been aired during segments of time, the annotated events comprising metadata instances that have semantic information about the events;
generate mappings between the content items and the annotated events, and between the content items and a plurality of topics;
determine a first affinity score between a first population of the users who authored content items mapped to a first one of the events or a first one of the topics and a second population of the users who authored content items mapped to a second one of the events or a second one of the topics;
determine a second affinity score between a third population of the users who authored content items mapped to a third one of the events or a third one of the topics and a fourth population of the users who authored content items mapped to a fourth one of the events or a fourth one of the topics;
determine a third affinity score between the second population and the fourth population; and
infer a fourth affinity score between the first population and third population based on the first, second, and third affinity scores.
11. The medium of claim 10 , wherein generating the mappings comprises
performing a feature extraction process on pairs including one of the content items and one of the events or one of the topics, the feature extraction generating content features that identify co-occurring textual information between the content item and the metadata instances of the annotated event or one or more keywords associated with the topic of the pair; and
generating scores for the content features indicating whether the content items refers to the annotated events or topics using a feature specific sub-function for the content feature.
12. The medium of claim 10 , wherein the instructions further cause the processor to:
access programming information regarding the events of time based broadcast media; and
annotate the events using an alignment function to estimate the likelihood that the events are described by the metadata instances.
13. The medium of claim 10 , wherein determining one of the affinity scores comprises:
determining an intersecting population of the users between one of the populations mapped to one of the events or topics, and another one of the populations mapped to the one of the events or topics;
determining the affinity score based on the intersecting population.
14. The medium of claim 13 , wherein determining one of the affinity scores comprises calculating a weighted sum over the users in the intersecting population.
15. The medium of claim 14 ,
wherein a weight is assigned to each user in the intersecting population, and
wherein each of the weights is determined based on at least one from the group consisting of:
a number of times the user has authored one of the content items associated with one of the mappings, and
a probability that the user will author a content item relevant to either the event or topic associated with one of the populations forming the intersecting population.
16. The medium of claim 14 , wherein determining the affinity score comprises normalizing the weighted sum by an average probability of users to author content items regarding the topics or events associated with intersecting population.
17. The medium of claim 14 , wherein determining the affinity score comprises normalizing the affinity score based on a number of users in the populations forming the intersecting population.
18. The medium of claim 17 , wherein determining the affinity score comprises normalizing the affinity score based on at least one from the group consisting of:
a cardinality of the union between the users of the populations forming the intersecting population and
a product of the cardinality between the populations forming the intersecting population.
19. A system comprising:
a computer server comprising a processor and a database, the computer server configured to:
access a plurality of content items authored by users of a social networking system;
access a plurality of annotated events of time based broadcast media, the events having been aired during segments of time, the annotated events comprising metadata instances that have semantic information about the events;
generate mappings between the content items and the annotated events, and between the content items and a plurality of topics;
determine a first affinity score between a first population of the users who authored content items mapped to a first one of the events or a first one of the topics and a second population of the users who authored content items mapped to a second one of the events or a second one of the topics;
determine a second affinity score between a third population of the users who authored content items mapped to a third one of the events or a third one of the topics and a fourth population of the users who authored content items mapped to a fourth one of the events or a fourth one of the topics;
determine a third affinity score between the second population and the fourth population; and
infer a fourth affinity score between the first population and third population based on the first, second, and third affinity scores.
20. The system of claim 19 , wherein generating the mappings comprises
performing a feature extraction process on pairs including one of the content items and one of the events or one of the topics, the feature extraction generating content features that identify co-occurring textual information between the content item and the metadata instances of the annotated event or one or more keywords associated with the topic of the pair; and
generating scores for the content features indicating whether the content items refers to the annotated events or topics using a feature specific sub-function for the content feature.
21. The system of claim 19 , wherein the computer server is further configured to:
access programming information regarding the events of time based broadcast media; and
annotate the events using an alignment function to estimate the likelihood that the events are described by the metadata instances.
22. The system of claim 19 , wherein determining one of the affinity scores comprises:
determining an intersecting population of the users between one of the populations mapped to one of the events or topics, and another one of the populations mapped to the one of the events or topics;
determining the affinity score based on the intersecting population.
23. The system of claim 22 , wherein determining one of the affinity scores comprises calculating a weighted sum over the users in the intersecting population.
24. The system of claim 23 ,
wherein a weight is assigned to each user in the intersecting population, and
wherein each of the weights is determined based on at least one from the group consisting of:
a number of times the user has authored one of the content items associated with one of the mappings, and
a probability that the user will author a content item relevant to either the event or topic associated with one of the populations forming the intersecting population.
25. The system of claim 23 , wherein determining the affinity score comprises normalizing the weighted sum by an average probability of users to author content items regarding the topics or events associated with intersecting population.
26. The system of claim 23 , wherein determining the affinity score comprises normalizing the affinity score based on a number of users in the populations forming the intersecting population.
27. The system of claim 26 , wherein determining the affinity score comprises normalizing the affinity score based on at least one from the group consisting of:
a cardinality of the union between the users of the populations forming the intersecting population and
a product of the cardinality between the populations forming the intersecting population.