Predicting and ranking search query results
View Patent ↗Techniques are described herein for providing search results that are ranked based on a predictor that predicts, for each of a number of objects, likelihoods that each particular object will be selected at different positions on a results list.
1. A machine-implemented method, comprising:
for a first object of a plurality of objects that are determined to be relevant to a first search query, determining a first predictor metric based at least in part on previous selections of the first object under circumstances in which the first object was at a first position in one or more first ordered lists of objects generated in response to one or more first search requests based on the first search query;
for a second object of the plurality of objects, determining a second predictor metric based at least in part on previous selections of the second object under circumstances in which the second object was at a second position in the one or more first ordered lists;
wherein each predictor metric reflects a likelihood that the associated object will be selected at one or more positions within second ordered lists of objects that are (a) determined to be relevant to the first search query and (b) ordered differently than said first ordered lists;
generating a ranked list by ranking the plurality of objects based on previous selection-based predictor metrics determined for the first and second objects, wherein the second object is placed at a position higher than the second position in the ranked list, and the first object is placed at a position lower than the first position in the ranked list; and
in response to receiving a second search request based on said first search query from a user, providing at least a portion of the ranked list to said user;
wherein the method is performed by one or more computing devices.
2. The method of claim 1 , wherein determining a predictor metric for an object is based at least in part on established selections of the object at at least one position in two or more first ordered lists of objects that match the search query.
3. The method of claim 1 , wherein determining a predictor metric for an object is based at least in part on established selections of objects that are not objects that match the first search query.
4. The method of claim 1 , wherein at least a portion of the objects that match the first search query represent image files.
5. The method of claim 1 , wherein the at least a portion of objects that match the first search query represent video files.
6. A machine-implemented method, comprising:
for each particular object of a plurality of objects that are determined to be relevant to a search query, determining a predictor for the particular object based at least in part on previous selections of the particular object under circumstances in which the particular object was at a particular position in one or more first ordered lists of objects that match the search query,
wherein said predictor reflects a likelihood that the particular object will be selected at various positions within second ordered lists of objects that are (a) determined to be relevant to the search query and (b) ordered differently than said first ordered lists;
generating a ranked list by ranking the plurality of objects based on previous selection-based predictors determined for each object in the plurality of objects; and
in response to receiving said search query from a user, providing at least a portion of the ranked list to said user;
wherein the predictor:
a) determines a performance metric for the particular object at the current position;
b) determines a second position for which to predict performance;
c) determines an expected performance metric based at least in part on historical selections of objects that have:
1) moved to the second position from a previous position that matches the current position of the particular object; and
2) previously had a performance metric matching the performance metric of the particular object;
wherein the method is performed by one or more computing devices.
7. The method of claim 6 , wherein determining a performance metric comprises:
dividing the actual percentage of selections by an expected percentage of selections.
8. The method of claim 7 , wherein the expected percentage of selections comprises of static data.
9. A computer-implemented method comprising:
determining a plurality of different orderings of a first of search results;
determining, for a first ordering of the plurality of orderings, a selection probability that reflects a probability that search results that are ordered according to the first ordering will be selected by users;
determining, for a second ordering of the plurality of orderings, a selection probability that reflects a probability that search results that are ordered according to the second ordering will be selected by users;
wherein the selection probability for each particular ordering is based on the number of predicted selections for each object in that ordering;
selecting, the first ordering in response to determining that the first ordering is associated with a highest selection probability among the selection probabilities that are associated with the first and second orderings;
wherein the object occupying the highest position in the first ordering is expected to receive less selections than another object at the first position in the second ordering; and
storing the selected ordering on a computer-readable storage medium;
wherein the method is performed by one or more computing devices.
10. The method of claim 9 , wherein the selection probability is based at least in part on information concerning numbers of previous selections of the search results under circumstances in which the search results were ordered according to the particular ordering.
11. A volatile or non-volatile non-transitory computer-readable storage medium storing instructions which, when executed by one or more processors, causes the one or more processors to perform the method recited in claim 1 .
12. A volatile or non-volatile non-transitory computer-readable storage medium storing instructions which, when executed by one or more processors, causes the one or more processors to perform the method recited in claim 2 .
13. A volatile or non-volatile non-transitory computer-readable storage medium storing instructions which, when executed by one or more processors, causes the one or more processors to perform the method recited in claim 3 .
14. A volatile or non-volatile non-transitory computer-readable storage medium storing instructions which, when executed by one or more processors, causes the one or more processors to perform the method recited in claim 4 .
15. A volatile or non-volatile non-transitory computer-readable storage medium storing instructions which, when executed by one or more processors, causes the one or more processors to perform the method recited in claim 5 .
16. A volatile or non-volatile non-transitory computer-readable storage medium storing instructions which, when executed by one or more processors, causes the one or more processors to perform the method recited in claim 6 .
17. A volatile or non-volatile non-transitory computer-readable storage medium storing instructions which, when executed by one or more processors, causes the one or more processors to perform the method recited in claim 7 .
18. A volatile or non-volatile non-transitory computer-readable storage medium storing instructions which, when executed by one or more processors, causes the one or more processors to perform the method recited in claim 8 .
19. A volatile or non-volatile non-transitory computer-readable storage medium storing instructions which, when executed by one or more processors, causes the one or more processors to perform the method recited in claim 9 .
20. A volatile or non-volatile non-transitory computer-readable storage medium storing instructions which, when executed by one or more processors, causes the one or more processors to perform the method recited in claim 10 .