Google Inc.

At+t Corp.

Lucent Technologies Inc.

Daryl Pregibon

Diane L Tang

Daniel Wright

Diane Lambert

Corinna Cortes

Ashish Agarwal

Michael M Meyer

Ilia Mirkin

James Angus Morrison

Eric Bauer Arbanovella

Susan Shannon

Theodore Donald Johnson

Thaddeus Julius Kowalski

Linda A Clark

Christopher T Volinsky

William H DuMouchel

A system of content/query feature based data structure retrieval of predicted values is provided. The system can create a data structure having a plurality of rows corresponding to individual content/query features and a plurality of columns corresponding to individual predicted values. The processors can obtain a set of session features associated with a selection by a computing device in response to a query, and a set of content/query features associated with the selection of the content item. The processors can retrieve, from the data structure, a set of predicted values for each of the set of content/query features. The processors can generate, for each of the set of content/query features, a set of aggregate predicted values for each of the set of content/query features, and can include the set of aggregate predicted values in the data structure.

Query feature based data structure retrieval of predicted values

User queries are received, with each query requesting a service from a server. Overlapping experiments are performed on at least a portion of the queries, with each experiment modifying one or more parameters associated with the queries or parameters associated with processing of the queries, and with the experiments organized into layers. Two or more experiments in different layers are allowed to be performed on the same query, and for any given layer, at most one experiment is allowed to be performed on the same query.

Overlapping experiments

A system obtains a first parameter (QP) associated with a quality of an advertisement among multiple advertisements, where the first quality parameter (QP) does not include a click through rate (CTR). The system functionally combines the first quality parameter (QP) with at least one other parameter and uses the functional combination to filter, rank or promote the advertisement among the multiple advertisements.

Using estimated ad qualities for ad filtering, ranking and promotion

A system obtains ratings associated with a first set of advertisements hosted by one or more servers, where the ratings indicate a quality of the first set of advertisements. The system observes multiple different first user actions associated with user selection of advertisements of the first set of advertisements and derives a statistical model using the observed first user actions and the obtained ratings. The system further observes second user actions associated with user selection of a second advertisement hosted by the one or more servers and uses the statistical model and the second user actions to estimate a quality of the second advertisement.

Estimating ad quality from observed user behavior

Apparatus and method for summarizing an original large data set with a representative data set. The data elements in both the original data set and the representative data set have the same variables, but there are significantly fewer data elements in the representative data set. Each data element in the representative data set has an associated weight, representing the degree of compression. There are three steps for constructing the representative data set. First, the original data elements are partitioned into separate bins. Second, moments of the data elements partitioned in each bin are calculated. Finally, the representative data set is generated by finding data elements and associated weights having substantially the same moments as the original data set.

Method and system for squashing a large data set

A method provides for mining information from large volumes of data regarding transactions. The method provides for inferring a behavioral characteristic of a party to the transaction based on a large volume of data concerning a multitude of parties. That inferred characteristic may be dynamic in nature.

Method for inferring behavioral characteristics based on a large volume of data

A cluster interface is generated which represents similarities of semantics between segments of code with respect to both the physical constructs of the code and the underlying operations performed by the code. The generated interface represents one or more code segments. Code segments to be analyzed are received by a computer system. Statistical internal information is extracted from each code segment. An external metric is generated which is based on the extracted statistical information. An interface display is created from the external metric which represents similarity relationships between the inputted code segments based on the extracted statistical information.

Growing community of inventors

Summit, NJ, United States of America

Daryl Pregibon