Microsoft Technology Licensing, LLC

Christopher J C Burges

John Carlton Platt

Robert L Rounthwaite

Dengyong Zhou

Ryen William White

Jonathan David Goldstein

Eric David Brill

Allison P Heath

Daniel F Plastina

Erin L Renshaw

Tao Tao

Qiang Wu

Henrique Sarmento Malvar

Cormac E Herley

Krysta Marie Svore

Arungunram Chandrasekaran Surendran

Geoffrey G Zweig

Hagai Thomas Attias

Elbio Renato Torres Abib

Bhuvan Middha

Somsak Sukittanon

Tal Shaked

Oliver B Downs

Aspects of the subject matter described herein relate to predicting and using search engine switching behavior. In aspects, switching components receive a representation of user interactions with at least one browser. The switching components derive information from the representation that is useful in predicting whether a user will switch search engines. The derived information and information about a user's current interaction with a browser is then used by a switch predictor to predict whether the user will switch search engines. This prediction may be used in a variety of ways examples of which are given herein.

Technologies for encouraging search engine switching based on behavior patterns

The subject disclosure is directed towards automated processes for generating sentence completion questions based at least in part on a language model. Using the language model, a sentence is located, and alternates for a focus word (or words) in the sentence are automatically provided. Also described is automated filtering candidate sentences to locate the sentence, filtering the alternates based upon elimination criteria, scoring sentences with the correct word and as modified the alternates, and ranking the alternates. Manual selection may be used along with the automated processes.

Generating sentence completion questions

Described is a technology for identifying sample data items (e.g., documents corresponding to query-URL pairs) having the greatest likelihood of being mislabeled when previously judged, and selecting those data items for re-judging. In one aspect, lambda gradient scores (information associated with ranked sample data items that indicates a relative direction and how 'strongly' to move each data item for lowering a ranking cost) are summed for pairs of sample data items to compute re-judgment scores for each of those sample data items. The re-judgment scores indicate a relative likelihood of mislabeling. Once the selected sample data items are re-judged, a new training set is available, whereby a new ranker may be trained.

Identification of sample data items for re-judging

A spam detection system is disclosed. The system includes a classifier training component that receives a first set of training pages labeled as normal pages and a second set of training pages labeled as spam pages. The training component trains a web page classifier based on both the first set of training pages and the second set of training pages. A spam detector then receives unlabeled web pages uses the web page classifier to classify the unlabeled web pages as spam pages or normal pages.

Link spam detection using smooth classification function

Described herein is a system that includes a receiver component that receives first scores for training points and second scores for the training points, wherein the first scores are individually assigned to the training points by a first ranker component and the second scores are individually assigned to the training points by a second ranker component. The apparatus further includes a determiner component in communication with the receiver component that automatically outputs a value for a parameter &#x3b1; based at least in part upon the first scores and the second scores, wherein &#x3b1; is used to linearly combine the first ranker component and the second ranker component.

Linear combination of rankers

A system described herein includes a trainer component that receives an estimated gradient of cost that corresponds to a first ranker component with respect to at least one training point and at least one query. The trainer component builds a second ranker component based at least in part upon the received estimated gradient. The system further includes a combiner component that linearly combines the first ranker component and the second ranker component.

Boosting a ranker for improved ranking accuracy

Predicting and using search engine switching behavior

A system described herein includes analyzer component that analyzes queries submitted by users and corresponding URLs selected by the users, wherein the queries include a first query and a second query, and wherein the analyzer component determines that the first query and the second query are substantially similar queries. The system additionally includes a correlator component that, responsive to the analyzer component determining that the first query and the second query are substantially similar, generates correlation data that indicates that the first and second queries are substantially similar.

Substantially similar queries

A collection of web pages is considered as a directed graph in which the pages themselves are nodes and the hyperlinks between the pages are directed edges in the graph. A trusted entity identifies training examples for spam pages and normal pages. A random walk is conducted through the directed graph that includes the collection of web pages and the stationary probabilities, and transitional probabilities, among the nodes in the directed graph are obtained. A classifier training component estimates a classification function that changes slowly on densely connected subgraphs within the directed graph. The classification function assigns a value to each of the nodes in the directed graph and identifies them as spam or normal pages based upon whether the value meets a given function threshold value.

A general probabilistic formulation referred to as 'Conditional Harmonic Mixing' is provided, in which links between classification nodes are directed, a conditional probability matrix is associated with each link, and where the numbers of classes can vary from node to node. A posterior class probability at each node is updated by minimizing a divergence between its distribution and that predicted by its neighbors. For arbitrary graphs, as long as each unlabeled point is reachable from at least one training point, a solution generally always exists, is unique, and can be found by solving a sparse linear system iteratively. In one aspect, an automated data classification system is provided. The system includes a data set having at least one labeled category node in the data set. A semi-supervised learning component employs directed arcs to determine the label of at least one other unlabeled category node in the data set.

Leveraging unlabeled data with a probabilistic graphical model

A dependency structure is used to divide samples corresponding to items to be ranked into leaf nodes, based on the rank of the items. The dependency structure is trained by splitting or merging training data received at given nodes based on selected features and selected thresholds for those features. A metric is then calculated which is indicative of performance of the node, in splitting the data. The trained structure is then used during runtime to rank items.

Ranking items by optimizing ranking cost function

Information about media objects within media streams is inferred based on repeat instances of the media objects within the media streams. A system and methods enable the monitoring of one or more media streams and the identification of repeat instances of media objects (e.g., audio and/or video objects) within the media streams. The monitoring and object repeat identification is performed by one or more server computers on a network. Information about a media object can be inferred based on repeat instances of the media object and based on repeat instances of related media objects. The information is transferred from a server to a client in response to a user query entered at the client through an interactive user interface.

Inferring information about media stream objects

A collection of web pages is modeled as a directed graph, in which the nodes of the graph are the web pages and directed edges are hyperlinks. Web pages can also be represented by content, or by other features, to obtain a similarity graph over the web pages, where nodes again denote the web pages and the links or edges between each pair of nodes is weighted by a corresponding similarity between those two nodes. A random walk is defined for each graph, and a mixture of the random walks is obtained for the set of graphs. The collection of web pages is then analyzed based on the mixture to obtain a web page analysis result. The web page analysis result can be, for example, clustering of the web pages to discover web communities, classifying or categorizing the web pages, or spam detection indicating whether a given web page is spam or content.

Growing community of inventors

Bellevue, WA, United States of America

Christopher J C Burges