Microsoft Technology Licensing, LLC

Quantarium Group, LLC

Quantarium, LLC

Microsoft Corporation

Mihai Petriuc

Dmitriy Meyerzon

Clement Ifrim

Adrian Fanaru

Nicolai Bodd

Andrei T Aron

Malcolm Cannon

Stefan Isbein Larimore

Vladimir Tankovich

Raif Khassanov

Artsiom Kokhan

Cezar Ungureanasu

Chadd Creighton Merrigan

John Smintina

Eugene Shnitko

Roxana Paval

Diana Manzat

Petru Cociorva

Michael James Taylor

Hugo Zaragoza

Dominic Suciu

Stephen Robertson

Petru Iulian Cociorva

Ioan Smintina

Vignesh Ram Nithin Kappagantula

Elena Petriuc

Andrei Aron

Systems and methods are disclosed for computer vision property evaluation. In certain embodiments, a method may comprise executing a computer vision property evaluation operation via a computing system. The computer vision property evaluation operation may include identifying an image of a selected property feature using a first neural network (NN) of the computing system, cropping the image of the selected property feature in a selected way to produce a cropped image using a second NN of the computing system, generating a categorization of the cropped image based on identified details of the selected property feature, and generating a classification of a property corresponding to the selected property feature based on the categorization.

Computer vision property evaluation

Systems and methods are disclosed for smart mapping, including dynamically recalculating a heatmap of displayed map entries based on changes to a map displayed on a viewport. In certain embodiments, a method may comprise providing a display of a map on a user interface, determining selected geographic tiles depicted on the display, and loading selected data corresponding to map entries from the selected geographic tiles only. The method may further comprise generating a heatmap on the display based on comparing the selected data, determining changes at the user interface resulting in a different set of geographic tiles depicted on the display, and dynamically generating a new heatmap based on the changes.

Smart mapping display adjustment

Smart mapping

This disclosure describes methods and systems for searching documents in a multi-tenant hosting environment. According to embodiments, to conserve hardware resources, a plurality of documents associated with a plurality of tenants may be mapped to the same search index in the multi-tenant hosting environment. In order to search documents associated only with a single tenant in the multi-tenant hosting environment, a tenant identifier is prepended to every key stored in the search index that is associated with the plurality of documents of the single tenant. Moreover, where one document links to another document within the multi-tenant hosting environment, the link is stored in a web graph when a source tenant identifier matches a target tenant identifier for the link. According to embodiments, when conducting a search, the link is resolved only if the link is stored in the web graph.

Tenantization of search result ranking

A search index structure which extends a typical composite index by incorporating an index which is optimized for fast retrieval from storage and which eliminates data which is specific to phrase searching. Other data is represented in a manner which allows it to be calculated rather than stored. Associating variable length entries with logical categories allows their length to be inferred from the category rather than stored. Using delta values between document IDs rather than the ID itself generates a compact, dense symbol set which is efficiently compressed by Huffman encoding or a similar compression method. Using an upper threshold to remove large, and thus rare, delta values from the symbol set prior to encoding further improves the encoding performance.

Search index format optimizations

Real estate ranking computation is calculated to sort real estate properties. Such computations use available information regarding real estate properties in any local markets and may help the real estate buyer to identify real estate properties with various ranked financial values. Suitably, a relatively small geographic area can be used based on the real estate buyer's specified criteria, and the real estate ranking computation is able to handle interactions among predictor variables, possesses suitable predictive confidence, and includes the capability for dynamically adjusting the underlying ranking computation as new patterns of real estate market emerge over time.

Ranking real estate based on its value and other factors

Technologies are described herein for providing a more efficient approach to ranking search results. An illustrative technology reduces an amount of ranking data analyzed at query time. In the technology, a term is selected, at index time, from a master index. The term corresponds to a number of documents greater than a threshold. A set of documents that includes the term is selected based on the master index. A rank is determined for each document in the set of documents that contains the term. Each document in the set of documents that contains the term is assigned to a top document list or a bottom document list based on the rank. Predefined values of at least part of the rank are stored in the top document list for documents in the top document list and are not stored in the bottom document list for documents in the bottom document list.

Index optimization for ranking using a linear model

Technologies are described herein for providing a more efficient approach to ranking search results. One method reduces an amount of ranking data analyzed at query time. In the method, a term is selected, at index time, from a master index. The term corresponds to a number of documents greater than a threshold. A set of documents that includes the term is selected based on the master index. A rank is determined for each document in the set of documents that contains the term. Each document in the set of documents that contains the term is assigned to a high ranking index or a low ranking index based on the simple rank.

An efficient determination of a click distance value is made for each document in a corpus of documents from data included in a locally-stored inverted index. The click distance is measurement of the number clicks or user navigations from a first document on the network to another document. Specialized words are included in the locally-stored inverted index. The specialized words relate source documents to a set of target documents. A click distance is assigned to a source document when an inverted index is queried for the corresponding set of target documents according to a query that passes in one of the specialized words. The process is repeated for each document in the corpus of documents.

Click distance determination

A query is separated into subqueries including a first subquery containing terms applicable to a first data store and a second subquery containing terms applicable to a second data store, where both data stores maintain information regarding the documents. Applying the first subquery to the first data store retrieves a first list of document identifiers associated with documents that satisfy the terms of the first subquery. The first list is combined with the second subquery to form a masked subquery, which is applied to the second data store. The masked subquery only seeks to identify document identifiers that both are included the first list and that satisfy terms of the second query. The document identifiers included in the first list may be ordered to match an order in which the document identifiers are ordered in the second data store.

Batching document identifiers for result trimming

A field-weighted search combines statistical information for each term across document fields in a suitably weighted fashion. Both field-specific term frequencies and field and document lengths are considered to obtain a field-weighted document weight for each query term. Each field-weighted document weight can then be combined in order to generate a field-weighted document score that is responsive to the overall query.

Field weighting in text searching

Growing community of inventors

Redmond, WA, United States of America

Mihai Petriuc