Google Inc.

Xuerui Wang

Rahul Sharma

Xiaodan Song

Daniel Li

Jie Han

Gaofeng Zhao

Xing Wei

Feng Li

Generating granular clusters for real-time processing is provided. The systems can identify tokens based on aggregating input from computing devices over a time interval. The systems can identify, based on metrics, a subset of tokens for cluster generation. The systems can generate, via a clustering technique, token clusters from the subset of the tokens, each of the token clusters comprising two or more tokens from the subset of the tokens. The systems can apply a de-duplication technique to each of the token clusters. The systems can apply a filtering technique to the token clusters to remove tokens erroneously grouped in a token cluster. The systems can assign, based on a selection process, a label for each of the token clusters. The systems can activate, based on a number of remaining tokens in each of the token clusters, a subset of the token clusters for real-time content selection.

Granular cluster generation for real-time processing

Machine learning based automatic audience segment in ad targeting

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for clustering data elements. In one aspect, a method includes determining a respective linkage value for each of multiple cluster pairs, where each cluster pair includes a respective first cluster and a respective second cluster. Determining a linkage value for a cluster pair includes determining a set of pairwise similarity values for the cluster pair. Each pairwise similarity value defines a similarity measure between: (i) a particular data element from the first cluster of the cluster pair, and (ii) a given data element from the second cluster of the cluster pair. The linkage value for the cluster pair is assigned as a given percentile of the set of pairwise similarity values, wherein the given percentile is greater than 0 and less than 100. A cluster pair is merged based on the linkage values of the cluster pairs.

Percentile linkage clustering

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for scoring criteria for content items. In one aspect, a method includes identifying a primary ranking signal and a set of auxiliary ranking signals for ranking a set of criteria for a content item. A primary score and a set of auxiliary scores can be identified for each particular criterion in the set of criteria. Each auxiliary score can be adjusted to generate adjusted auxiliary scores. The adjusting can include applying, to at least a portion of the auxiliary scores, a transformation function that reduces an amount of skewness among the auxiliary scores. A ranking score can be determined for each particular criterion based on a function of the primary score for the particular criterion and the adjusted auxiliary scores.

Scoring criteria for a content item

Methods, systems, and apparatus for diversifying selection criteria. An example method includes accessing a first set of selection criteria for a content item; selecting, based on the first set of selection criteria, a second set of selection criteria, the second set of selection criteria being a proper subset of the first set of selection criteria, the selection of each of the criteria in the second set comprising: determining a collective performance measure based on criteria that have been selected from the first set of selection criteria for inclusion in the second set of selection criteria, determining, for each criterion that has not been selected from the first set, a performance contribution to the collective performance measure, and selecting from the first set a selection criterion with the highest performance contribution relative to the performance contributions of other selection criteria that have not been selected for inclusion in the second set.

Growing community of inventors

Palo Alto, CA, United States of America

Xuerui Wang