Back to results
US 8,078,617 B1Granted
Model based ad targeting
AssigneeGoogle Inc.
InventorsWilliam T. Neveitt, Ralph U. Gasser
Life of this patent
Estimating the term…
Abstract
System and method for quickly selecting content items (e.g., documents) having highest ranking scores in response to a query for content items. Base scores for a plurality of content items are determined based on attributes of the content items. One or more attributes are extracted from the query, the base scores of at least some of the content items are adjusted based on relationships between the query attributes and the content item attributes. A subset of the plurality of content items having the highest adjusted scores are selected, and the selected subset of content items are output.
Claims33
A computer-implemented method comprising:
- determining base scores for a plurality of content items based on attributes of the content items;
- receiving a query;
- deriving one or more attributes from the query;
- adjusting the base scores of at least some of the content items based on relationships between the query attributes and the content item attributes, the adjustments to the base scores determined at least in part according to a model that includes a first mapping relationship that maps query keys to content item keys and a second mapping relationship that maps content item keys to content items, the query keys representing features of queries and the content item keys representing features of content items;
- selecting a subset of the plurality of content items having the highest adjusted scores;
- and
- outputting the selected subset of content items.
4 more elements
A computer-implemented method comprising:
- generating a first map that links query keys to content item keys, the query keys representing features of queries, the content item keys representing features of content items;
- generating a second map that links content item keys to content items to indicate which content items are associated with each content item feature;
- and
- storing the first and second maps in a storage.
1 more element
A computer-implemented method comprising:
- receiving a query,
- identifying a subset of content items having highest scores among a set of content items based on information about features of the query, a first predetermined mapping relationship between query features and content item features, and a second predetermined mapping relationship between content item features and content items, by computing adjustments to pre-computed baseline scores for content items having features that are affected by the features of the received query according to the first and second mapping relationships;
- and
- outputting the subset of content items having the highest scores.
1 more element
A computer-implemented method comprising:
- training an odds model over a training set of queries and content items, the odds model comprising a query map that has information about how features of queries affect features of content items, each query feature being associated with one or more content item features, and a content item map that has information about which content items in the training set of content items are associated with each content item feature;
- and
- outputting the odds model.
A computer-implemented method comprising:
- receiving a query,
- identifying a first subset of content items having highest approximate scores among a set of content items based on information about features of the query and an approximate odds model, the approximate odds model comprising a first predetermined mapping relationship between query features and content item features, and a second predetermined mapping relationship between content item features and content items;
- applying an exact odds model to determine exact scores of the first subset of content items;
- and
- outputting a second subset of content items having highest exact scores among the first subset of content items.
2 more elements
A system, comprising:
- a storage to store scores of a plurality of content items in which the scores are determined based on features of the content items;
- and
- a processing device to:
- receive a query,
- extract one or more features from the query,
- adjust the scores of at least some of the content items based on relationships between the query features and the content item features, the adjustments to the base scores determined at least in part according to a model that includes a first mapping relationship that maps query keys to content item keys and a second mapping relationship that maps content item keys to content items, the query keys representing features of queries and the content item keys representing features of content items, and
- select a subset of the plurality of content items having the highest adjusted scores.
4 more elements
A system comprising:
- means for storing scores of a plurality of content items in which the scores are determined based on attributes of the content items;
- and
- means for receiving a query,
- extracting one or more attributes from the query,
- adjusting the scores of at least some of the content items based on relationships between the query attributes and the content item attributes, the adjustments to the base scores determined at least in part according to a model that includes a first mapping relationship that maps query keys to content item keys and a second mapping relationship that maps content item keys to content items, the query keys representing features of queries and the content item keys representing features of content items, and
- selecting a number of content items having the highest adjusted scores from the plurality of content items.
3 more elements
Dependencies are inferred from the claim text.