Google Inc.

Shankar R Ponnekanti

Jia Liu

Deepak Ravichandran

Yijian Bai

Sittichai Jiampojamarn

Manojav Patil

Dong Lin

Oren Eli Zamir

Ting Liu

Nicholas Johnson

Xiaonan Zhang

Methods, systems, and apparatus, including computer programs encoded on a computer-readable storage medium, for serving content to a user. A method includes: receiving a request for content from a user, the content to be displayed in a slot associated with an online resource; determining an expected activity score for the user based at least in part on one or more criteria associated with a context of presenting the online resource to the user; comparing the expected activity score for the user to a threshold; selecting one or more content items to serve to the user based at least in part on the comparison; and serving the selected content items to the user responsive to the request.

Expected activity of a user

Systems and methods for predicting content performance with interest data include receiving a content selection request that includes a client identifier. One or more topical interest categories associated with the client identifier may be used as inputs to a prediction model to predict the likelihood of an online action occurring as a result of third-party content being selected. The predicted likelihood may be used to select third-party content.

Network based content transmission based on client device parameters

A computerized method and apparatus for evaluating content on a computer network. The method includes obtaining a quality score of content configured for display with a web page, wherein the quality score is based at least in part on keywords associated with the content and either a search query or metadata associated with the web page. The method also includes identifying a user metric of a computing device associated with the search query or the metadata. The method further includes generating an adjusted quality score of the content based on the quality score and the user metric. The method also includes selecting a parameter for an auction based on the adjusted quality score, wherein the parameter indicates a relation between a bid value based auction and a content quality based auction.

Content evaluation based on users browsing history

Predicting content performance with interest data

Content evaluation based on user's browsing history

A computerized method and system operable to build a device identifier similarity model with online event signals and determine similar network device identifiers. A processing circuit receives a first set of network device identifiers. The processing circuit represents each network device identifier of the first set by feature data associated with each network device identifier's network activity, where the feature data is associated with the content clicked-on or converted-on. The processing circuit applies abstractions on the feature data to form concepts. The processing circuit derives at least one hierarchy of feature data based on the keywords and concepts of the feature data. The processing circuit expands the feature data based on the derived at least one hierarchy of feature data and generates the device identifier similarity model based on the expanded feature data. The processing circuit is also capable of determining long-term and short-term history events.

Device identifier similarity models derived from online event signals

A computerized method for generating a representative model for a plurality of different models identified by similar feature data. A processing circuit receives a plurality of different models, each model configured for use in generating a second set of network user identifiers based on a first set of network user identifiers. The processing circuit receives feature data for each of the plurality of different models, each feature data having a corresponding feature weight data. The processing circuit identifies similar models within the plurality of different models based on a similarity of the feature data between models within the plurality of different models. The processing circuit generates the representative model to represent the similar models. The representative model may be used to generate the second set of network user identifiers based on the feature data and corresponding weight data of the representative model.

Generating a representative model for a plurality of models identified by similar feature data

A computerized method of creating a classifier model to determine whether a network user should be added to a list of similar network users. A first list of network users, a second list of network users who are not on the first list of network users and characteristic data associated with the network users' network activity are received. A time period is designated. A first category of network activities performed by a network user on the first list within the designated time period is identified. A second category of network activities performed at any time by network users on the second list is identified. A third category of network activities performed by network users on the first list at a time outside of the designated time period is identified. The classifier model is generated based on the first, second and third categories of network activities.

Creating a classifier model to determine whether a network user should be added to a list

A computerized method, system for, and computer-readable medium operable to select an algorithm for generating models configured to identify similar user identifiers. A first plurality of models generated by a first algorithm is received. A plurality of lists of similar user identifiers is generated. User queries associated with user identifiers on the plurality of lists of similar user identifiers are identified. Predicted click-through rates for the user queries is received. An average predicted click-through rate is computed for each model based on the predicted click-through rates. A weighted average predicted click-through rate associated with the first plurality of models is computed. The weighted average predicted click-through rate for the first plurality of models can be compared to a weighted average predicted click-through rate for a second plurality of models generated by a second algorithm. The algorithm for generating models is selected based on the comparison.

Selecting an algorithm for identifying similar user identifiers based on predicted click-through-rate

A method of setting a threshold similarity score value for a first plurality of network user identifiers. The first plurality of network user identifiers, a second plurality of network user identifiers and characteristic data associated with the network user identifiers is received. A performance target and an experimental threshold similarity score value are designated. A similarity score between the first and second plurality of network user identifiers is calculated. Performance statistics data for each of the second plurality of network user identifiers having a similarity score greater than or equal to the experimental threshold similarity score value is collected and compared to the similarity score of the network user identifier. Based on the comparison, the experimental threshold similarity score value is adjusted to a similarity score value that achieves the performance target and the threshold similarity score value is set to the adjusted experimental threshold similarity score value.

Adjust similar users identification based on performance feedback

Systems and methods of selecting consumers to receive content on a computer network are provided. A user list identifying a first plurality of users having a group of features corresponding to internet activity of the first plurality of users can be obtained at a computing device. A subgroup of features can be selected from the group of features, and a cluster of users of the first plurality of users can be identified. The users of the cluster of users can each have at least one feature of the subgroup of features. A supplemental user having a supplemental feature can be identified. A correlation between the supplemental feature and at least one feature of the subgroup of features can be determined, and an expanded user list that includes at least one of the first plurality of users and the supplemental user can be generated.

Growing community of inventors

Sunnyvale, CA, United States of America

Shankar R Ponnekanti