Link localization by country
Described are systems and methods that solve localization problems using Machine Learning models to compute country vectors for each linked content item and present content items in response to requests based on the country vectors. For example, a request from a user in Country A may be processed to determine candidate content items responsive to the request and to determine Country A as the country corresponding to the request. The candidate content items may then be processed to determine, for each candidate content item, a country vector corresponding to Country A as indicative of the relevance of the content item to Country A. Content items that are more likely than not to be relevant to the country of the request (e.g., Country A), as indicated by the respective country vector, may be considered as responsive and all other candidate content items discarded.
1. A method comprising:
determining a country associated with a request for content items;
determining, based at least in part on the request, a plurality of candidate content items responsive to the request, each of the plurality of candidate content items being associated with a corresponding link;
computing, for each of the plurality of candidate content items and using a trained machine learning system, a corresponding country vector;
comparing each corresponding country vector against a threshold;
determining that a first corresponding country vector of the corresponding country vectors exceeds the threshold; and
returning, in response to the request, the corresponding link associated with a first content item corresponding to the first corresponding country vector.
2. The method of claim 1 , further comprising:
determining that a second corresponding country vector from the corresponding country vectors does not exceed the threshold; and
discarding, based on the determination that the second corresponding country vector does not exceed the threshold, a second content item corresponding to the second corresponding country vector.
3. The method of claim 1 , wherein computing each corresponding country vector for each of the plurality of the candidate content items is based at least in part on each corresponding link.
4. The method of claim 1 , wherein computing the corresponding country vector for each of the plurality of content items includes at least one of:
determining a historical user preference corresponding to a historical access to the content item; or
determining a link locale corresponding to a link of the content item.
5. The method of claim 3 , wherein computing each corresponding country vector is based at least in part on each corresponding link includes retracting a link path of the corresponding link.
6. A computing system, comprising:
one or more processors; and
a memory storing program instructions that, when executed by the one or more processors, cause the one or more processors to at least:
receive a request;
determine a location corresponding to the request;
for each of a plurality of content items determined to be responsive to the request and having a corresponding link:
compute, using a trained machine learning system, a location vector indicative of a first relevance of the content item to the location; and
compare the location vector against a threshold to determine if the location vector exceeds the threshold; and
return, as responsive to the request, at least one content item of the plurality of content items having a respective location vector that exceeds the threshold.
7. The computing system of claim 6 , wherein determining the location corresponding to the request includes determining at least one of a language corresponding to the request, a location of a device associated with the request, or a user country preference.
8. The computing system of claim 6 , wherein the program instructions, when executed by the one or more processors, further cause the one or more processors to determine at least one of:
a language corresponding to a content item of the plurality of content items;
a historical user preference of the content item;
a link locale for the content item; or
third party data corresponding to the content item.
9. The computing system of claim 8 , wherein the program instructions, when executed by the one or more processors to cause the one or more processors to compute the location vector, further cause the one or more processors to at least:
compute the location vector based at least in part on one or more of the language, the historical user preference, or the link locale.
10. The computing system of claim 6 , wherein the program instructions, when executed by the one or more processors, further cause the one or more processors to at least:
determine a plurality of locations corresponding to the request; and
for each of the plurality of content items:
compute, for each respective location of the plurality of locations, a second location vector indicative of a second relevance of the content item to the respective location.
11. The computing system of claim 6 , wherein:
computing the location vector for each of the plurality of content items is based at least in part on the corresponding link.
12. The computing system of claim 11 , wherein computing the second location vector for each of the plurality of content items is based at least in part on each corresponding link and includes retracting a link path of the corresponding link.
13. The computing system of claim 6 , wherein the program instructions, when executed by the one or more processors, further cause the one or more processors to at least:
discard at least one content item of the plurality of content items having a corresponding location vector that does not exceed the threshold.
14. The computing system of claim 6 , wherein the program instructions, when executed by the one or more processors, further cause the one or more processors to at least:
rank the plurality of content items based at least in part on one or more of a second relevance to the request or the location vector corresponding to each content item.
15. The computing system of claim 6 , wherein the program instructions, when executed by the one or more processors, further cause the one or more processors to at least:
maintain, in a data store, the location vectors computed for each of content items.
16. A computer-implemented method, comprising:
determining a first location corresponding to a request for content items from a user;
for each of a first plurality of content items determined to be responsive to the request and includes a corresponding link, computing, using a trained machine learning system, at least one location vector, the at least one location vector indicative of a relevance of the content item to the first location;
comparing the at least one location vector for each of the first plurality of content items to a threshold to determine from the first plurality of content items, a second plurality of content items having a corresponding location vector value that exceeds the threshold, wherein the second plurality of content items is less than the first plurality of content items; and
providing access to at least a portion of the second plurality of content items to the user.
17. The computer-implemented method of claim 16 , wherein the first location is determined based on one or more of a language corresponding to the request, a second location of a device from which the request was received, a third location of a user, or a user country preference.
18. The computer-implemented method of claim 16 , wherein computing the at least one location vector includes:
determining, for the content item, at least one of a language of the content item, a language of a link to the content item, a historical user preference corresponding to the content item, a link locale for the content item, or third party data corresponding to the content item;
providing, as input to the machine learning system, at least one of the language of the content item, the language of the link to the content item, the historical user preference corresponding to the content item, the link locale for the content item, or the third party data corresponding to the content item; and
receiving, from the machine learning system and in response to the input, the at least one location vector for the content item.
19. The computer-implemented method of claim 16 , further comprising:
ranking the at least a portion of the second plurality of content items to generate a ranked list of the second plurality of content items,
wherein providing access includes providing, for presentation to the user, the ranked list.