Aol Inc.

Facebook, Inc.

Microsoft Technology Licensing, LLC

America Online, Inc.

Mercury Kingdom Assets Limited

Bright Sun Technologies

Aol Llc, a Delaware Limited Company

Aleksander R Kolcz

Joshua Alspector

Abdur Chowdhury

David Dolan Lewis

Steven Michael Beitzel

Eliot Carleton Gillum

Krishna C Vitaldevara

Mihai Costea

Jason D Walter

Douglas John Hines

Wen-tau Yih

Ryan Charles Colvin

Kevin Andrew Chan

Robert L McCann

Chad W Mills

Harsh Satyanarayan Dangayach

Yuxiang Xu

A query phrase may be automatically classified to one or more topics of interest (e.g., categories) to assist in routing the query phrase to one or more appropriate backend databases. A selectional preference query classification technique may be used to classify the query phrase based on a comparison between the query phrase and patterns of query phrases. Additionally, or alternatively, a combination of query classification techniques may be used to classify the query phrase. Topical classification of a query phrase also may be used to assist a search system in delivering auxiliary information to a user who entered the query phrase. Advertisements, for instance, may be tailored based on classification rather than query keywords.

Web query classification

In general, a two or more stage spam filtering system is used to filter spam in an e-mail system. One stage includes a global e-mail classifier that classifies e-mail as it enters the e-mail system. The parameters of the global e-mail classifier generally may be determined by the policies of e-mail system owner and generally are set to only classify as spam those e-mails that are likely to be considered spam by a significant number of users of the e-mail system. Another stage includes personal e-mail classifiers at the individual mailboxes of the e-mail system users. The parameters of the personal e-mail classifiers generally are set by the users through retraining, such that the personal e-mail classifiers are refined to track the subjective perceptions of their respective user as to what e-mails are spam e-mails. Retraining data for the personal e-mail classifiers may be aggregated and a subset of the aggregate may be chosen for use in retraining the global e-mail classifier.

Online adaptive filtering of messages

A classification system includes a signature-based duplicate detector and an inductive classifier that share attribute information. To perform the duplicate detection and the classification, the duplicate detector and inductive classifier are first initialized by generating a lexicon of attributes for the duplicate detector and a classification model for the classifier. To develop a classification model, a training set of documents of known class are used by the classifier to determine the attributes of the documents that are most useful in classifying an unknown document. The model is developed from these attributes. Attribute information containing the attributes determined by the classifier is then passed to the duplicate detector and the duplicate detector uses the attribute information to generate the lexicon of attributes.

Method for classifying unknown electronic documents based upon at least one classificaton

Campaign detection techniques are described. In implementations, a signature is computed for each of a plurality of emails to be communicated by a service provider to respective intended recipients. A determination is made that two or more of the plurality of emails is similar based on the respective signatures. Responsive to a finding that a number of similar emails exceeds a threshold, an indication is output that the similar emails have a likelihood of being involved in a spam campaign.

Campaign detection

In general, a two or more stage spam filtering system is used to filter spam in an e-mail system. One stage includes a global e-mail classifier that classifies e-mail as it enters the e-mail system. The parameters of the global e-mail classifier generally may be determined by the policies of e-mail system owner and generally are set to only classify as spam those e-mails that are likely to be considered spam by a significant number of users of the e-mail system. Another stage includes personal e-mail classifiers at the individual mailboxes of the e-mail system users. The parameters of the personal e-mail classifiers generally are set by the users through retraining, such that the personal e-mail classifiers are refined to track the subjective perceptions of their respective user as to what e-mails are spam e-mails.

In a single-signature duplicate document system, a secondary set of attributes is used in addition to a primary set of attributes so as to improve the precision of the system. When the projection of a document onto the primary set of attributes is below a threshold, then a secondary set of attributes is used to supplement the primary lexicon so that the projection is above the threshold.

Duplicate document detection

Simplifying lexicon creation in hybrid duplicate detection and inductive classifier systems

Systems and methods are provided for ranking documents. In accordance with certain implementations, a user selects one or more documents for viewing from a set of documents. Documents selected by the user for viewing within a predetermined time period are maintained in a profile associated with the user. At least one positive word vector is generated using words contained in at least a segment of at least one of the documents stored in the profile. At least one negative word vector is generated using words contained in at least a segment of at least one of the documents that was not selected by the user for viewing. Document word vectors are generated for documents to be ranked. A vector space relationship analysis of the positive word vector, the negative word vector, and the document word vectors is performed. The documents are ranked based on the performed vector space relationship analysis.

Filtering system for providing personalized information in the absence of negative data

An e-mail filter is used to classify received e-mails so that some of the classes may be filtered, blocked, or marked. The e-mail filter may include a classifier that can classify an e-mail as belonging to a particular class and an e-mail grouper that can detect substantially similar, but possibly not identical, e-mails. The e-mail grouper determines groups of substantially similar e-mails in an incoming e-mail stream. For each group, the classifier determines whether one or more test e-mails from the group belongs to the particular class. The classifier then designates the class to which the other e-mails in the group belong based on the results for the test e-mails.

Group based spam classification

Reliability of duplicate document detection algorithms

Detecting spam from metafeatures of an email message. As a part of detecting spam, the email message is accessed and a distribution of numerical values is accorded to a set of features of the email message. It is determined whether the distribution of numerical values accorded the set of features of the email message is consistent with that of spam. Access is provided to the determination of whether the email message has a distribution of numerical values accorded the set of features that is consistent with that of spam.

Detecting spam from metafeatures of an email message

Systems and methods are provided for personalizing advertising for a user. In accordance with certain implementations, information is accessed indicating which documents were selected by a user and which documents were not selected by a user. At least one positive word vector is generated using words contained in at least one of the selected documents, and at least one negative word vector is generated using words contained in at least one of the unselected documents. Document word vectors are generated, and a document rank order is established based on a vector space relationship analysis. Categories associated with the documents are ranked based on the document rank order, and the ranked categories are sent to an ad server. Advertising material associated with the ranked categories may then be received from the ad server in a selected context.

In one aspect, a data item is input into a scoring classifier such that the scoring classifier indicates that the data item belongs to a first class. A determination is made as to the amount of retraining of the scoring classifier, based on the data item, that is required to cause the scoring classifier to indicate that the data item belongs to a second class. A reliability measure is determined based on the required amount of retraining and a class of the data item is determined based, at least in part, on the reliability measure.

Growing community of inventors

Kirkland, WA, United States of America

Aleksander R Kolcz