Method and system for ranking search results based on category demand normalized using impressions
Described herein are methods and systems for promoting item listings that satisfy a query based on the item listings being assigned to certain categories that have, based on historical click data, exhibited high demand characteristics for the query. Consistent with some embodiments, a certain number of leaf-level categories are identified based on demand data for those categories, and the item listings assigned to those categories are promoted through a normalized weighting factor derived in part based on the click probability score associated with the category, clicks per impression rate, and weighted clicks per impression by ranking rate.
1. A computer-implemented method comprising:
processing a query, by a hardware processor of a machine, to identify a set of item listings, each item listing associated with an item or service being offered and assigned to a leaf-level category;
identifying a leaf-level category for each item listing satisfying the query;
obtaining for the query a click probability score for each leaf-level category to which an item listing satisfying the query has been assigned;
identifying up to a predetermined number of leaf-level categories from all leaf-level categories from the set of item listings identified with the query with click probability scores exceeding a same threshold score for all leaf-level categories;
for each of the identified leaf-level categories, calculating a category boost score for use in determining the order in which the item listings are to be presented in a search results page;
normalizing the category boost score for one or more identified leaf-level categories; and
presenting a search results page with the item listings ordered based in part on the normalized category boost score for the leaf-level category to which each item listing is assigned.
2. The computer-implemented method of claim 1 , wherein normalizing further comprises:
calculating a constant boost score for the one or more identified leaf-level categories.
3. The computer-implemented method of claim 1 , wherein normalizing further comprises:
determining a respective number of clicks per impression for each leaf-level category; and
calculating the normalized category boost score of a leaf-level category based on the respective number of clicks per impression.
4. The computer-implemented method of claim 1 , wherein normalizing further comprises:
determining a weighted impression by rank for each leaf-level category;
determining a respective weighted number of clicks per impression by rank for each leaf-level category; and
calculating the normalized category boost score of each leaf-level category based on the respective weighted number of clicks per impression by rank.
5. The computer-implemented method of claim 1 , wherein the threshold score is derived as a percentage of the click probability score of the leaf-level category with a highest click probability score.
6. The computer-implemented method of claim 1 , wherein the threshold score is derived by dividing the click probability score of the leaf-level category with a highest click probability score by one less than the predetermined number.
7. The computer-implemented method of claim 1 , wherein the category boost score for each identified leaf-level category is derived based in part on the click probability score of each identified leaf-level category.
8. The computer-implemented method of claim 1 , wherein the item listings are ordered based on a ranking score derived with an algorithm utilizing the category boost score as a factor.
9. The computer-implemented method of claim 1 , wherein the click probability score for each category represents a probability, for a particular query, that an item listing assigned to the category will be selected from a search results page, the click probability score for each category derived based on analysis of historical click data.
10. The computer-implemented method of claim 1 , wherein obtaining a click probability score for each leaf-level category to which an item listing satisfying the query has been assigned includes dividing a number of clicks for a particular leaf-level category by the total number of clicks for all leaf-level categories to which an item listing satisfying the query has been assigned.
11. A system for an item listing presentation management, the system comprising:
at least one processor comprising:
a listing identifier module configured to process a query to identify a set of item listings, each item listing associated with an item or service being offered and assigned to a leaf-level category, and to identify a leaf-level category for each item listing satisfying the query;
a probability score module configured to obtain for the query a click probability score for each leaf-level category to which an item listing satisfying the query has been assigned, and identifying up to a predetermined number of leaf-level categories from all leaf-level categories from the set of item listings identified with the query with click probability scores exceeding a same threshold score for all leaf-level categories;
a category boost module configured to calculate a category boost score, for each of the identified leaf-level categories, for use in determining the order in which the item listings are to be presented in a search results page;
a normalizing module, implemented with the least one processor, configured to normalize the category boost score for one or more identified leaf-level categories; and
a listing generator module configured to present a search results page with the item listings ordered based in part on the normalized category boost score for the leaf-level category to which each item listing is assigned.
12. The system of claim 11 , wherein the normalizing module comprises:
a constant boost module configured to calculate a constant boost score for the one or more identified leaf-level categories.
13. The system of claim 11 , wherein the normalizing module comprises:
a click through rate module configured to determine a respective number of clicks per impression for each leaf-level category, and to calculate the normalized category boost score of a leaf-level category based on the respective number of clicks per impression.
14. The system of claim 11 , wherein the normalizing module comprises:
a weight click through rate by rank module configured to determine a weighted impression by rank for each leaf-level category, to determine a respective weighted number of clicks per impression by rank for each leaf-level category, and to calculate the normalized category boost score of each leaf-level category based on the respective weighted number of clicks per impression by rank.
15. The system of claim 11 , wherein the item listing presentation management module is to derive the threshold score as a percentage of the click probability score of the leaf-level category with a highest click probability score.
16. The system of claim 11 , wherein the item listing presentation management module is to derive the threshold score by dividing the click probability score of the leaf-level category with a highest click probability score by one less than the predetermined number.
17. The system of claim 11 , wherein the item listing presentation management module is to derive the category boost score for each identified leaf-level category based in part on the click probability score of each identified leaf-level category.
18. The system of claim 11 , wherein the item listing presentation management module is to ordered the item listings based on a ranking score derived with an algorithm utilizing the category boost score as a factor.
19. The system of claim 11 , wherein the click probability score for each category represents a probability, for a particular query, that an item listing assigned to the category will be selected from a search results page, the click probability score for each category derived based on analysis of historical click data.
20. The system of claim 11 , wherein obtaining a click probability score for each leaf-level category to which an item listing satisfying the query has been assigned includes dividing a number of clicks for a particular leaf-level category by the total number of clicks for all leaf-level categories to which an item listing satisfying the query has been assigned.
21. A non-transitory computer-readable storage medium storing a set of instructions that, when executed by a processor, cause the processor to perform operations, comprising:
processing a query to identify a set of item listings, each item listing associated with an item or service being offered and assigned to a leaf-level category;
identifying a leaf-level category for each item listing satisfying the query;
obtaining for the query a click probability score for each leaf-level category to which an item listing satisfying the query has been assigned;
identifying up to a predetermined number of leaf-level categories from all leaf-level categories from the set of item listings identified with the query with click probability scores exceeding a same threshold score for all leaf-level categories;
for each of the identified leaf-level categories, calculating a category boost score for use in determining the order in which the item listings are to be presented in a search results page;
normalizing the category boost score for one or more identified leaf-level categories; and
presenting a search results page with the item listings ordered based in part on the normalized category boost score for the leaf-level category to which each item listing is assigned.