US 8,527,507 B2Granted

Custom ranking model schema

AssigneeMicrosoft Corporation

InventorsYauhen Shnitko, Karen Beattie Massey, Dmitriy Meyerzon, Victor Poznanski

Life of this patent

Estimating the term…

Loading patent…

Abstract

A customizable ranking model of a search engine using custom ranking model configuration and parameters of a pre-defined human-readable format. The architecture can employ a markup language schema to represent the custom ranking model. In one implementation, the schema developed utilizes XML (extensible markup language) for representing the custom ranking model. Weights for dynamic and static relevance ingredients can be altered per ranking model and new relevance ingredients can be added. Additionally, features are provided for improving relevance such as adding terms to a thesaurus for synonym expansion, for example, the ability to deal with single terms either as compounds, and/or using custom word breaking rules.

Claims18

Dependent claims

A computer for providing a customizable ranking model, comprising:

a processor;
the processor executing a process, wherein the processor is adapted to:
rank documents resulting from a search through the customizable ranking model;
use context including language in the customizable ranking model to rank the documents;
manage single terms in the context as compounds using custom word breaking rules;
add single terms to a thesaurus for synonym expansion in a query;
utilizing a locale identifier number for the customizable ranking model to represent a combination of:
the language, a region, and a sort for the documents associated with a date, a time, a number, a currency, a formatting, a calendar preference, an input method, and a sorting preference of the documents;
operating the language as a query dependent feature;
represent the customizable ranking model through a schema written in a human-readable code;
and
employ a root level ranking model element with a query dependent features element and a query independent features element as child elements as a structure for the elements, wherein the query dependent features element has a child query dependent feature element including property identification, name, weight and length normalization attributes, and the query independent features element has a category feature element, a query independent feature element and a language feature element.

9 more elements

A computer for providing a customizable ranking model, comprising:

a processor;
the processor executing a process, wherein the process is adapted to:
employ a text ranking formula for processing multiple text fields through the customizable ranking model;
tune the customizable ranking model according to a language, a vertical, and a site experience;
calculate scores using the text ranking formula according to a function of features defined in the customizable ranking model;
add single terms to a thesaurus for synonym expansion in a query;
determine term frequency in the text ranking formula utilizing number of times a term appears in a property of documents including the text fields while evaluating a length of the property as a number of terms, an average length of the property in a collection in the documents, a number of the documents in a corpus, the number of the documents containing the term, a weight of a field, and a weight of length normalization;
represent elements of the customizable ranking model ranked according to the scores in XML through a schema;
and
employ a root level ranking model element with a query dependent features element and a query independent features element as child elements as a structure for the elements, wherein the query dependent features element has a child query dependent feature element including property identification, name, weight and length normalization attributes, and the query independent features element has a category feature element, a query independent feature element and a language feature element.

7 more elements

A computer-implemented ranking model method, comprising:

defining a schema of configuration information and parameters for a customizable ranking model according to a markup language;
tuning the customizable ranking model according to a language, a vertical, and a site experience;
operating the language as a query independent feature;
utilizing context including the language in the customizable ranking model;
manage single terms in the context as compounds using custom word breaking rules;
adding single terms to a thesaurus for synonym expansion of the single terms in a query;
calculating scores using a formula according to a function of features defined in the customizable ranking model;
determining term frequency in a text ranking formula utilizing number of times the single terms appear in property of documents while evaluating a length of the property as a number of the single terms, an average length of the property in a collection in the documents, a number of the documents in a corpus, a number of documents containing the single terms, a weight of a field, and a weight of length normalization;
utilizing a locale identifier number for the customizable ranking model to represent a combination of:
the language, a region, and a sort for documents associated with at a date, a time, a number, a currency, a formatting, a calendar preference, an input method, and a sorting preference of the context of the documents;
employing a root level ranking model element with a query dependent features element and a query independent features element as child elements as a structure for the elements, wherein the query dependent features element has a child query dependent feature element including property identification, name, weight and length normalization attributes, and the query independent features element has a category feature element, a query independent feature element and a language feature element;
and
configuring the customizable ranking model according to the schema sorted according to the scores.

10 more elements

Dependencies are inferred from the claim text.