Google Inc.

Microsoft Technology Licensing, LLC

Hartmut Maennel

Dmitry Korolev

Krzysztof Duleba

Matthias Heiler

Luuk Van Dijk

Thomas Hofmann

Aileen Tang

Jeffrey David Oldham

Michael Schaer

Zoltan Szamonek

Fiona Herring

Wojciech Gajewski

Justyna Sidorska

Stephen Wharton Provine

Habib Heydarian

Dmity Korolev

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for improving geographic targeting of digital content. In some implementations, a targeting request that identifies a target geographic region is received. Groups of geographic regions that each include the target geographic region and at least another geographic region are identified. Combined targeting accuracies are computed for the groups of geographic regions. One or more of the groups of geographic regions are selected based on their combined targeting accuracies being higher than a targeting accuracy for the target geographic region. Data describing the selected one or more groups of geographic regions is provided for output in response to the targeting request.

Targeting of digital content to geographic regions

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for improving geographic targeting of digital content. One of the methods includes receiving a targeting request that identifies a target geographic region; identifying one or more groups of geographic regions that each include at least two geographic regions, including the target geographic region, wherein the one or more groups of geographic regions are identified based on respective combined targeting accuracies, the respective combined targeting accuracy of each of the one or more groups being higher than a targeting accuracy for the target geographic region; and providing data describing the identified one or more groups of geographic regions in response to the targeting request.

Methods, systems, and apparatus, including computer program products, for constructing text classifiers. The method includes receiving a collection of candidate phrases for a given topic; filtering the received candidate phrases to remove erroneously included candidate phrases; assigning weights to the candidate phrases including scoring each candidate phrase using an initial classifier and assigning weights to the candidate phrases based on the scores; and generating a linear classifier using the filtered and weighted candidate phrases, where the linear classifier varies the weights for each phrase candidate depending on the length of the document being classified.

Construction of text classifiers

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for determining geographic locations of devices. One of the methods includes obtaining an estimated user location associated with each respective IP address block based on observed events from the IP address block; obtaining an estimate of a probability model p(ev|loc), the probability model p(ev|loc) including a respective probability distribution of interest locations for each of multiple user locations; wherein obtaining the estimate of the probability model p(ev|loc) includes calculating p(ev|loc) from a p(zone|loc) matrix and a p(ev|zone) matrix; and using the estimate for the probability model p(ev|loc) and the observed events to calculate an estimate for multiple probability distributions X(loc) associated with a respective IP address block.

Refining location detection from a query stream

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for determining geographic locations. One of the methods includes obtaining a sequence of events, each of the events including geographical location information, from a first device to be located; determining, for each event and each of a plurality of geographical locations, a probability that the respective event was obtained from a second device given that the second device is located at the respective geographical location; determining a probability that the sequence of events was obtained from the second device, including using a model representing how sequences of events are generated by network devices; and determining for each of the plurality of geographical locations a probability that the first device is located at the respective geographical location using the probability that the sequence of events was obtained.

Determining geo-locations of users from user activities

Embodiments of the disclosed subject matter construct linear combinations of a set of defined 'basic' features for each pair of sample and known media files. The linear combination of those features that give the optimal correlation coefficients (i.e., most-correlated) is then found. The highest correlation coefficients then indicate the 'best' features to use in detecting a match of the sample file.

Identifying features for media file comparison

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for classifying resources using scores from multiple classifiers. In general, one aspect of the subject matter described in this specification can be embodied in methods that include the actions of receiving identifying a collection of documents to classify; receiving a plurality of classifiers for scoring a document with respect to a specified property; for each document in the collection, applying each of the plurality of classifiers, each classifier generating a score associated with a likelihood that the document has the specified property, combining the scores from each classifier including applying a multiple classifier model that uses monotonic regression to combine the plurality of classifiers, and classifying the document as having the specified property based on the combined score.

Classifying documents using multiple classifiers

Various technologies and techniques are disclosed for using Petri Nets for diagnostics of concurrent programs. One or more initial traces are obtained from a concurrent program that executed properly. The initial trace is used to create a Petri Net. To generate the Petri Net, unique task identifiers and places of interest are generated, and transitions are added. The resulting Petri Net is stored for later use in differential debugging. A trace is obtained from a later execution of the concurrent program that did not execute properly. The later trace is run against the Petri Net to identify at least one difference. A Petri Net can also be used for model checking of concurrent programs. A model checking process is performed to analyze possible executions in the Petri Net to look for one or more problems. A result of the model checking process is output.

Constructing petri nets from traces for diagnostics

A method for automatically retrieving source code during software debugging sessions using a development system includes stopping in an object with compiled code where the compiled code is associated with corresponding source code via a debugging support file. The compiled code can be run on different hosts. During a debugging session, a local debugger portion of the software in the development system requests a unique resource locator for the source code. A remote software debugger portion on a host responds by providing a hosttype, which specifies the host specific search algorithm, and the original path for the source code. A development system host access component requests a file with matching name and original path from the host. An original source file checksum is compared with a retrieved checksum. Upon a match, the retrieved source code is stored in the development system and accessed for user display.

Growing community of inventors

Zurich, Switzerland

Hartmut Maennel