Yahoo! Inc.

Gordon Sun

Zhaohui Zheng

Hongyuan Zha

Ya Zhang

Michael E Palmer

Keke Chen

The system and method of the present invention allows for the determination of the relevance of a content item to a query through the use of a machine learned relevance function that incorporate query differentiation. A method for selecting a relevance function to determine a relevance of a query-content item pair comprises generating a training set comprising one or more content item-query pairs. Content item-query pairs in the training set are collectively used to determine the relevance function by minimizing a loss function according to a relevance score adjustment function that accounts for query differentiation. The monotocity of relevance score adjustment function allows the trained relevance function to be directly applied to new queries.

Learning retrieval functions incorporating query differentiation for information retrieval

Embodiments of the present invention provide for methods, systems and computer program products for learning ranking functions to determine the ranking of one or more content items that are responsive to a query. The present invention includes generating one or more training sets comprising one or more content item-query pairs, determining preference data for the one or more query-content item pairs of the one or more training sets and determining labeled data for the one or more query-content item pairs of the one or more training sets. A ranking function is determined based upon the preference data and the labeled data for the one or more content-item query pairs of the one or more training sets. The ranking function is then stored for application to query-content item pairs not contained in the one or more training sets.

Learning ranking functions incorporating boosted ranking in a regression framework for information retrieval and ranking

Embodiments of the present invention provide for methods, systems and computer program products for learning ranking functions to determine the ranking of one or more content items that are responsive to a query. The present invention includes generating one or more training sets comprising one or more content item-query pairs and determining one or more contradicting pairs in a given training sets. An optimization function to minimize the number of contradicting pairs in the training set is formulated, and modified by incorporating a grade difference between one or more content items corresponding to the query in the training set and applied to each query in the training set. A ranking function is determined based on the application of regression trees on the queries of the training set minimized by the optimization function and stored for application to content item-query pairs not contained in the one or more training sets.

Learning ranking functions incorporating isotonic regression for information retrieval and ranking

The present invention relates to systems and methods for determining a content item relevance function. The method comprises collecting user preference data at a search provider for storage in a user preference data store and collecting expert-judgment data at the search provider for storage in an expert sample data store. A modeling module trains a base model through the use of the expert-judgment data and tunes the base model through the use of the user preference data to learn a set of one or more tuned models. A measure (B measure) is designed to evaluate the balanced performance of tuned model over expert judgment and user preference. The modeling module generates or selects the content item relevance function from the tuned models with B measure as the selection criterion.

[object Object]

System and method for the determination of the relevance of a content item to a query through the use of a machine learned relevance function that incorporates click-through features of the content items. A method for selecting a relevance function to determine a relevance of a query-content item pair comprises generating training set having one or more query-URL pairs labeled for relevance based on their click-through features. The labeled query-URL pairs are used to determine the relevance function by minimizing a loss function that accounts for click-through features of the content item. The computed relevance function is then applied to the click-through features of unlabeled content items to assign relevance scores thereto. An inverted click-through index of query-score pairs is formed and combined with the content index to improve relevance of search results.

System and method for indexing web content using click-through features

A method and apparatus are provided for determining when electronic documents stored in a large collection of documents are similar to one another. A plurality of similarity information is derived from the documents. The similarity information may be based on a variety of factors, including hyperlinks in the documents, text similarity, user click-through information, similarity in the titles of the documents or their location identifiers, and patterns of user viewing. The similarity information is fed to a combination function that synthesizes the various measures of similarity information into combined similarity information. Using the combined similarity information, an objective function is iteratively maximized in order to yield a generalized similarity value that expresses the similarity of particular pairs of documents. In an embodiment, the generalized similarity value is used to determine the proper category, among a taxonomy of categories in an index, cache or search system, into which certain documents belong.

Growing community of inventors

Redwood Shores, CA, United States of America

Gordon Sun