Identifying similar items based on global interaction history
One embodiment sets forth technique for computing a similarity score between two digital items is computed based on interaction histories associated with global users and interaction histories associated with local users. Global counts indicating the number of interactions associated with each unique pair of digital items are weighted based on a mixing rate. The weighted global counts are then combined with local counts to compute total counts. An effective interaction probability indicating the likelihood of a user interacting with one digital item in the pair of digital items after interacting with the other digital item in the pair is computed based on the total counts. The effective interaction probability is then corrected for noise, resulting in a similarity score indicating the similarity between the pair of digital items.
1. A computer-implemented method for identifying two or more similar items within a local catalogue, the method comprising:
generating, at a web server computing system, a web application interface that includes a plurality of interface elements, wherein each interface element included in the plurality of interface elements is individually selectable when the web application interface is displayed;
identifying, at the web server computing system, a different digital content item for each interface element included in the plurality of interface elements, wherein a given digital content item identified for a first interface element included in the plurality of interface elements is identified by:
determining that the web application interface is to be presented to a user that has interacted with a previously-presented digital content item included in a local catalogue,
determining a set of global counts based on interaction data associated with a global group of users, wherein a first global count in the set of global counts indicates a number of times the global group of users has interacted with a first digital content item in a global catalogue, and a second global count in the set of global counts indicates a number of times the global group of users has interacted with a second digital content item in the global catalogue after interacting with the first digital content item, and wherein the set of global counts is stored to a first section of an interaction history database,
determining a set of local counts based on interaction data associated with a local group of users, wherein a first local count in the set of local counts indicates a number of times the local group of users has interacted with the previously-presented digital content item, and a second local count in the set of local counts indicates a number of times the local group of users has interacted with the given digital content item in the local catalogue after interacting with the previously-presented digital content item, and wherein the previously-presented digital content item corresponds to the first digital content item and the given digital content item corresponds to the second digital content item, the local group of users interacting with the previously-presented and given digital content items via web content browsers interacting with the web server computing system through the network, wherein the set of local counts is stored to a second section of the interaction history database,
combining the set of global counts and the set of local counts to compute an interaction probability that indicates the likelihood of a particular user within the local group of users interacting with the given digital content item after interacting with the previously-presented digital content item,
computing a similarity score that indicates the similarity between the previously-presented digital content item and the given digital content item based on the interaction probability and a popularity score associated with the given digital content item, and
selecting the given digital content item for presentation in the first interface element based on the similarity score; and
transmitting, by the web server computing system to a content browser associated with the user, the web application interface that presents the given digital content item within the first interface element, wherein one or more interactions with the first interface element captured by the content browser are transmitted over the network to the web server computing system and recorded within the second section of the interaction history database.
2. The method of claim 1 , further comprising filtering the interaction data associated with the global group of users to remove data related to items that are not included in the local catalogue.
3. The method of claim 1 , wherein combining the set of global counts and the set of local counts comprises determining a mixing rate between the set of global counts and the set of local counts that indicates a weight applied to the interaction data associated with a global group of users.
4. The method of claim 3 , wherein combining the set of global counts and the set of local counts further comprises computing a global probability that indicates the likelihood that a global user within the global group of users would interact with the second item after interacting with the first item.
5. The method of claim 4 , wherein combining the set of global counts and the set of local counts further comprises computing a local probability that indicates the likelihood that a local user within the local group of users would interact with the given item after interacting with the previously-presented item.
6. The method of claim 5 , wherein combining the set of global counts and the set of local counts further comprises weighting the global probability according to the mixing rate, and then combining the local probability and the weighted global probability to compute the interaction probability.
7. The method of claim 1 , wherein the popularity score associated with the given item indicates the likelihood that the particular user would randomly interact with the given item.
8. The method of claim 7 , wherein computing the similarity score comprises computing a ratio of the interaction probability and the popularity score.
9. The method of claim 1 , further comprising applying a noise correction function to the similarity score to filter the data that was captured in the set of global counts or the set of local counts.
10. The method of claim 1 , wherein:
the local group of users comprises users in a first local geographic region; and
the global group of users comprises users not in the first local geographic region.
11. A non-transitory computer readable medium storing instructions that, when executed by a processor, cause the processor to identify two or more similar items within a local catalogue, by performing the steps of:
generating, at a web server computing system, a web application interface that includes a plurality of interface elements, wherein each interface element included in the plurality of interface elements is individually selectable when the web application interface is displayed;
identifying, at the web server computing system, a different digital content item for each interface element included in the plurality of interface elements, wherein a given digital content item identified for a first interface element included in the plurality of interface elements is identified by:
determining that the web application interface is to be presented to a user that has interacted with a previously-presented digital content item included in a local catalogue,
determining a set of global counts based on interaction data associated with a global group of users, wherein a first global count in the set of global counts indicates a number of times the global group of users has interacted with a first digital content item in a global catalogue, and a second global count in the set of global counts indicates a number of times the global group of users has interacted with a second digital content item in the global catalogue after interacting with the first digital content item, the global group of users interacting with the first and second digital content items via web content browsers interacting with the web server computing system through a network, wherein the set of global counts is stored to a first section of an interaction history database,
determining a set of local counts based on interaction data associated with a local group of users, wherein a first local count in the set of local counts indicates a number of times the local group of users has interacted with the previously-presented digital content item, and a second local count in the set of local counts indicates a number of times the local group of users has interacted with the given digital content item in the local catalogue after interacting with the previously-presented digital content item, and wherein the previously-presented digital content item corresponds to the first digital content item and the given digital content item corresponds to the second digital content item, and wherein the set of local counts is stored to a second section of the interaction history database,
combining the set of global counts and the set of local counts to compute an interaction probability that indicates the likelihood of a particular user within the local group of users interacting with the given digital content item after interacting with the previously-presented digital content item,
computing a similarity score that indicates the similarity between the previously-presented digital content item and the given digital content item based on the interaction probability and a popularity score associated with the given digital content item, and
selecting the given digital content item for presentation in the first interface element based on the similarity score; and
transmitting, by the web server computing system to a content browser associated with the user, the web application interface that presents the given digital content item within the first interface element, wherein one or more interactions with the first interface element captured by the content browser are transmitted over the network to the web server computing system and recorded within the second section of the interaction history database.
12. The non-transitory computer readable medium of claim 11 , further comprising filtering the interaction data associated with the global group of users to remove data related to items that are not included in the local catalogue.
13. The non-transitory computer readable medium of claim 11 , wherein combining the set of global counts and the set of local counts comprises determining a mixing rate between the set of global counts and the set of local counts that indicates a weight applied to the interaction data associated with a global group of users.
14. The non-transitory computer readable medium of claim 13 , wherein combining the set of global counts and the set of local counts further comprises computing a global probability that indicates the likelihood that a global user within the global group of users would interact with the second item after interacting with the first item.
15. The non-transitory computer readable medium of claim 14 , wherein combining the set of global counts and the set of local counts further comprises computing a local probability that indicates the likelihood that a local user within the local group of users would interact with the given item after interacting with the previously-presented item.
16. The non-transitory computer readable medium of claim 15 , wherein combining the set of global counts and the set of local counts further comprises weighting the global probability according to the mixing rate, and then combining the local probability and the weighted global probability to compute the interaction probability.
17. The non-transitory computer readable medium of claim 11 , wherein the popularity score associated with the given item indicates the likelihood that the particular user would randomly interact with the given item.
18. The non-transitory computer readable medium of claim 17 , wherein computing the similarity score comprises computing a ratio of the interaction probability and the popularity score.
19. The non-transitory computer readable medium of claim 11 , further comprising applying a noise correction function to the similarity score to filter the data that was captured in the set of global counts or the set of local counts.
20. A computer system, comprising:
a memory; and
a processor configured to:
generate a web application interface that includes a plurality of interface elements, wherein each interface element included in the plurality of interface elements is individually selectable when the web application interface is displayed;
identify a different digital content item for each interface element included in the plurality of interface elements, wherein a given digital content item identified for a first interface element included in the plurality of interface elements is identified by:
determining that the web application interface is to be presented to a user that has interacted with a previously-presented digital content item included in a local catalogue,
determine a set of global counts based on interaction data associated with a global group of users, wherein a first global count in the set of global counts indicates a number of times the global group of users has interacted with a first digital content item in a global catalogue, and a second global count in the set of global counts indicates a number of times the global group of users has interacted with a second digital content item in the global catalogue after interacting with the first digital content item, the global group of users interacting with the first and second digital content items via web content browsers interacting with the web server computing system through a network, wherein the set of global counts is stored to a first section of an interaction history database,
determine a set of local counts based on interaction data associated with a local group of users, wherein a first local count in the set of local counts indicates a number of times the local group of users has interacted with the previously-presented digital content item, and a second local count in the set of local counts indicates a number of times the local group of users has interacted with the given digital content item in the local catalogue after interacting with the previously-presented digital content item, and wherein the previously-presented digital content item corresponds to the first digital content item and the given digital content item corresponds to the second digital content item, and wherein the set of local counts is stored to a second section of the interaction history database,
combine the set of global counts and the set of local counts to compute an interaction probability that indicates the likelihood of a particular user within the local group of users interacting with the given digital content item after interacting with the previously-presented digital content item,
compute a similarity score that indicates the similarity between the previously-presented digital content item and the given digital content item based on the interaction probability and a popularity score associated with the given digital content item, and
select the given digital content item for presentation in the first interface element based on the similarity score; and
transmit, to a content browser associated with the user, the web application interface that presents the given digital content item within the first interface element, wherein one or more interactions with the first interface element captured by the content browser are transmitted over the network to the web server computing system and recorded within the second section of the interaction history database.
21. The computer system of claim 20 , wherein combining the set of global counts and the set of local counts comprises determining a mixing rate between the set of global counts and the set of local counts that indicates a weight applied to the interaction data associated with a global group of users.
22. The computer system of claim 20 , wherein:
the local group of users comprises users in a first local geographic region; and
the global group of users comprises users not in the first local geographic region.
23. The non-transitory computer readable medium of claim 11 , wherein:
the local group of users comprises users in a first local geographic region; and
the global group of users comprises users not in the first local geographic region.