Ranking of content based on implied relationships
The present technology has the ability to establish connections between content that do not have direct or explicit relationships. Implicit influence relationships can be established from user download sequence data, campaign data with keyword targeting, and content review data that mentions other content. Using these influence relationships, the relevance of content items can be determined based on the influence relationship of linked content items and a similarity relationship of content items. However, the importance of the influence relationship in ranking content items can vary depending on the parameters against which the content item is considered relevant. To address this, the present technology includes a context-driven factor that is used as a weight to adjust the impact of the influence relationship of the ranking, depending on the parameters against which the content item is considered relevant.
1 . A method for ranking content items by at least one processor and a memory, the method comprising:
identifying a plurality of result content items matching at least one criterion, the plurality of result content items identified from content items in a collection of content items; and
ranking the content items in the plurality of result content items based on a similarity relationship between the content items in the plurality of result content items, and an influence relationship between pairs of the content items in the plurality of result content items, wherein:
the similarity relationship between a first content item and a second content item is based on a comparison of embeddings of the first content item and the second content item in an embedding space derived, at least in part, from content metadata; and
the influence relationship between the first content item and the second content item is based on aggregated signals indicative that user engagement with the first content item increases a propensity for user engagement with the second content item.
2 . The method of claim 1 , further comprising:
determining the influence relationship between the pairs of content items in the collection of content items based on content item transitions, targeted campaign data, and review data.
3 . The method of claim 1 , wherein embeddings in the embedding space are derived from a title, a description, a user review, and a user rating of the content items.
4 . The method of claim 1 , further comprising:
constructing a graph of relationships between the content items in the collection of content items, wherein edges connecting the content items to other content items are defined by a relevance value that represents a relevance of the content items connected by the edges, the relevance value being made up of the similarity relationship and the influence relationship.
5 . The method of claim 1 , further comprising:
determining a context-driven factor from the plurality of result content items matching the at least one criterion, wherein when ranking the plurality of result content items, the similarity relationship is weighted by the context-driven factor and the influence relationship is weighted by an inverse of the context-driven factor.
6 . The method of claim 1 , wherein the content items are items of invitational content, the method further comprising:
receiving a request for an item of invitational content by a content delivery system, wherein the request includes the at least one criterion; and
sending a highly ranked result content item in response to the request.
7 . The method of claim 1 , further comprising:
providing a user interface to configure a targeted campaign, wherein a first item of invitational content that is configured to invite engagement with the first content item is to be targeted to be presented in association with at least one second content item;
recommending one or more highly ranked content items to be selected as the at least one second content item, wherein the at least one criterion is relevance to the first content item;
receiving a selection of the one or more highly ranked content items in the user interface; and
configuring the targeted campaign to serve the first item of invitational content in association with the selected one or more highly ranked content items.
8 . The method of claim 1 , wherein the aggregated signals include aggregated sequence data of content item transitions, wherein the first content item has a stronger influence relationship on the second content item when users transition to the second content item recently after interacting with the first content item more frequently than other content items in the aggregated sequence data.
9 . The method of claim 1 , wherein the aggregated signals include aggregated targeted campaign data, wherein the first content item has a stronger influence relationship on the second content item when targeted campaign data associated with the second content item targets users of the first content item more frequently than other content items in the aggregated targeted campaign data.
10 . The method of claim 1 , wherein the aggregated signals include aggregated review data, wherein the first content item has a stronger influence relationship on the second content item when first review data associated with the first content item references the second content item more frequently than other content items in the aggregated review data.
11 . The method of claim 1 , wherein the influence relationship indicates that a content item is influential when the content item at least one of is referred to by one or more highly influential content items or when the content item bridges discoveries of other content items.
12 . A system comprising:
at least one processor; and
a memory storing instructions that, when executed by the at least one processor, configures the system to:
identify a plurality of result content items matching at least one criterion, the plurality of result content items identified from content items in a collection of content items;
rank the content items in the plurality of result content items based on a similarity relationship between the content items in the plurality of result content items, and an influence relationship between pairs of the content items in the plurality of result content items, wherein:
the similarity relationship between a first content item and a second content item is based on a comparison of embeddings of the first content item and the second content item in an embedding space derived, at least in part, from content metadata; and
the influence relationship between the first content item and the second content item is based on aggregated signals indicative that user engagement with the first content item increases a propensity for user engagement with the second content item.
13 . The system of claim 12 , wherein the instructions further configure the system to:
determine the influence relationship between the first content item and the second content item in the collection of content items based on content item transitions, targeted campaign data, and review data.
14 . The system of claim 12 , wherein embeddings in the embedding space are derived from a title, a description, a user review, and a user rating of the content items.
15 . The system of claim 12 , wherein the instructions further configure the system to:
construct a graph of relationships between the content items in the collection of content items, wherein edges that connect the content items to other content items are defined by a relevance value that represents a relevance of the content items connected by the edges, the relevance value being made up of the similarity relationship and the influence relationship.
16 . The system of claim 12 , wherein the instructions further configure the system to:
determine a context-driven factor from the plurality of result content items matching the at least one criterion, wherein when ranking the result content items, the similarity relationship is weighted by the context-driven factor and the influence relationship is weighted by an inverse of the context-driven factor.
17 . A non-transitory computer-readable storage medium, the non-transitory computer-readable storage medium comprising instructions that when executed, configure at least one processor to:
identify a plurality of result content items matching at least one criterion, the plurality of result content items identified from content items in a collection of content items;
rank the content items in the plurality of result content items based on a similarity relationship between the content items in the plurality of result content items, and an influence relationship between pairs of the content items in the plurality of result content items, wherein
the similarity relationship between a first content item and a second content item is based on a comparison of embeddings of the first content item and the second content item in an embedding space derived, at least in part, from content metadata; and
the influence relationship between the first content item and the second content item is based on aggregated signals indicative that user engagement with the first content item increases a propensity for user engagement with the second content item.
18 . The non-transitory computer-readable storage medium of claim 17 , wherein the instructions further configure the at least one processor to:
determine the influence relationship between he first content item and the second content item in the collection of content items based on content item transitions, targeted campaign data, and review data.
19 . The non-transitory computer-readable storage medium of claim 17 , wherein embeddings in the embedding space are derived from a title, a description, a user review, and a user rating of the content items.
20 . The non-transitory computer-readable storage medium of claim 17 , wherein the instructions further configure the at least one processor to:
construct a graph of relationships between the content items in the collection of content items, wherein edges that connect the content items to other content items are defined by a relevance value that represents a relevance of the content items connected by the edges, the relevance value being made up of the similarity relationship and the influence relationship.
21 . The non-transitory computer-readable storage medium of claim 17 , wherein the instructions further configure the at least one processor to:
determine a context-driven factor from the plurality of result content items matching the at least one criterion, wherein when ranking the result content items, the similarity relationship is weighted by the context-driven factor and the influence relationship is weighted by an inverse of the context-driven factor.