Microsoft Technology Licensing, LLC

Kevyn Breca Collins-Thompson

Paul Nathan Bennett

Ryen William White

Susan Theresa Dumais

Ming Gang Liu

Daryl T Lawton

Sebastian De La Chica

David Alexander Sontag

David Maxwell Chickering

Karthik Raman

Daniel John Liebling

Milad Shokouhi

Charles B Schweizer

Siranush Sarkizova

Marc Sloan

Lidan Wang

David A Sontag

Paul Bennett

Various technologies related to generating and applying content retrieval rules are described herein. A content retrieval rule maps a combination of a query and a context to one of a query reformulation or content. The content retrieval rule is learned from search logs of a search engine, and is applied when the query having the context is received at the search engine.

Learning and using contextual content retrieval rules for query disambiguation

A request can be received and a request reading level representation for the request can be inferred. In response to the request, the request reading level representation can be compared with one or more reading difficulty level representations for one or more response items. Also in response to the request, one or more indications of results of comparing the request reading level representation with one or more reading difficulty level representations for the one or more response items can be returned. The indication(s) may include a ranking of the response items. The ranking can be based at least in part on a request reading level representation for the query and reading difficulty level representations for the response items. The response item(s) may also be returned.

Using reading levels in responding to requests

Various technologies described herein pertain to predicting intrinsically diverse sessions and retrieving information for such intrinsically diverse sessions. Search results retrieved by a search engine responsive to executing a query are received. A query classifier can be employed to determine whether the query is intrinsically diverse or not intrinsically diverse based on one or more features of the query and session interaction properties. The query is intrinsically diverse when included in an intrinsically diverse session directed towards a task, where the query and disparate queries included in the intrinsically diverse session are directed towards respective subtasks of the task. An objective function can be evaluated based at least upon the query to compute an optimized value when the query is determined to be intrinsically diverse. The search results can be presented on a display screen according to the optimized value when the query is determined to be intrinsically diverse.

Prediction and information retrieval for intrinsically diverse sessions

Technologies are described herein that pertain to optimizing a ranker component for a risk-oriented objective. Various definitions of risk are described herein, wherein risk is based upon variance in performance scores assigned to the ranker component for respective queries in a data store. Additionally, risk is optionally based upon variance in relative performance of the ranker component versus a baseline ranker component.

Optimizing a ranker for a risk-oriented objective

One or more systems and/or techniques are provided for constructing a query classification index that can be used to classify a query into relevant categories. Where documents in an index are classified into one or more category predictions for a category hierarchy, classification metadata is generated for categories to which a document in the index has been classified. Further, the classification metadata is associated to the corresponding documents in the index. Additionally, a query of the index can be classified using the metadata associated to the documents in the index, and query results can be provided that are classified by the one or more categories identified by the classification of the query.

Query classification

A query processing system is described herein for personalizing results for a particular user. The query processing system operates by receiving a query from a particular user u who intends to find results that satisfy the query with respect to a topic T, the user being characterized by user information &#x3b8;. In one implementation, the query processing system then produces a generic topic distribution Pr(T|q) associated with the query that is germane to a population of generic users, as well as a user-specific query-dependent topic distribution Pr(T|q,&#x3b8;) for the particular user. The query processing system then produces personalized results for the particular user based on Pr(T|q) and Pr(T|q,&#x3b8;). The query processing system can use multiple techniques to produce Pr(T|q,&#x3b8;), such as, in one approach, a discriminative learning approach.

Functionality for personalizing search results

Discriminating between documents scanned in a batch scanning process is achieved based on various analyses of the constituent document pages. The data provided by the various analyses are compared with each other to determine whether successive pages belong to the same document. Scanned documents result in a page sequence that is analyzed to extract one or more feature attributes for each page. The feature attributes are provided to a feature comparison process in order to assess the similarity of successive pages. If a sufficient likelihood of similarity is found, the compared pages are deemed to be from the same document; otherwise, they are deemed to be from different documents, indicating the existence of a document break. Based on the document breaks, separate scan files may be established. In this manner, the present invention represents eliminates the requirement of user intervention.

Method and apparatus for discriminating between documents in batch scanned document files

Discriminating between documents scanned in a batch scanning process is achieved based on various analyses of the constituent document pages. The data provided by the various analyses are compared with each other to determine whether successive pages belong to the same document. Scanned documents result in a page sequence. The page sequence is then analyzed to extract one or more features attributes for each page. The feature attributes are provided to a feature comparison process in order to assess the similarity of successive pages. If a sufficient likelihood of similarity is found, then the compared pages are deemed to be from the same document; otherwise, they are deemed to be from different documents, indicating the existence of a document break. Through the display of the page sequence, a user may optionally modify the location of one or more document breaks. Based on the document breaks, separate scan files may be established. In this manner, the present invention represents eliminates the requirement of user intervention.

Described is a system and method for improving string matching in a noisy channel environment. The invention provides a method for identifying string candidates and analyzing the probability that the string candidate matches a user-defined string. In one implementation, a find engine receives a query string, converts an image file into a textual file, and identifies each instance of the query string in the textual file. The find engine identifies candidates within the textual file that may match the query string. The find engine refers to a confusion table to help identify whether candidates that are near matches to the query string are actually matches to the query string but for a common recognition error. Candidates meeting a probability threshold are identified as matches to the query string. The invention further provides for analysis options including word heuristics, language models, and OCR confidences.

Growing community of inventors

Seattle, WA, United States of America

Kevyn Breca Collins-Thompson