USING COMPUTATIONAL ENGINES TO IMPROVE SEARCH RELEVANCE
First Claim
1. A method for providing query results in response to a user query, comprising using a computing device to perform steps for:
- receiving a current user query having multiple search terms;
evaluating the current user query to identify one or more subqueries, each subquery comprised of one or more of the search terms of the current user query;
evaluating each subquery to select one or more structured data sets from among a plurality of structured data sets from which to retrieve a partial answer corresponding each subquery;
for each partial answer retrieved from a selected structured data set, reformulating the current user query to construct a reformulated query by replacing the corresponding subquery of the current user query with the corresponding partial answer;
retrieving a set of ranked search results corresponding to reformulated query; and
presenting the set of ranked search results on a display device.
2 Assignments
0 Petitions
Accused Products
Abstract
An “Iterative Query Reformulator” provides various techniques for using a computational engine to reformulate initial queries through one or more iterations. This query reformulation process ensures that results returned from search engines or recommendation systems using a reformulated query have improved relevance relative to results that would have been returned using only the initial query. More specifically, the Iterative Query Reformulator provides an end to end solution that uses computations from one or more knowledge databases or knowledge sources to find “partial answers” to subqueries derived or extracted from an initial query. These partial answers are then used to reformulate the initial query, with the reformulated query being used by the search engines or recommendations systems to provide results that are highly relevant to the initial query. Determinations of whether to continue reformulation iterations are based on evaluating user metrics from historical search logs having queries that match reformulated queries.
69 Citations
20 Claims
-
1. A method for providing query results in response to a user query, comprising using a computing device to perform steps for:
-
receiving a current user query having multiple search terms; evaluating the current user query to identify one or more subqueries, each subquery comprised of one or more of the search terms of the current user query; evaluating each subquery to select one or more structured data sets from among a plurality of structured data sets from which to retrieve a partial answer corresponding each subquery; for each partial answer retrieved from a selected structured data set, reformulating the current user query to construct a reformulated query by replacing the corresponding subquery of the current user query with the corresponding partial answer; retrieving a set of ranked search results corresponding to reformulated query; and presenting the set of ranked search results on a display device. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10)
-
-
11. A system for improving relevance of search results, comprising:
-
a user input device for receiving a current query from a user; a user satisfaction measurement device for evaluating one or more user metrics within a set of historical search logs for determining a user satisfaction level for a set of query results corresponding to the current query; a computational engine device for iteratively reformulating the current query to construct a new current query, with such iterations repeating until the user satisfaction level exceeds a predetermined threshold, and wherein iteratively reformulating the current query comprises; evaluating the current query to identify one or more subqueries, each subquery comprised of one or more of the search terms of the current query, evaluating each subquery to select one or more structured data sets from among a plurality of structured data sets from which to retrieve a partial answer corresponding each subquery, for each partial answer retrieved from a selected structured data set, reformulating the current query to construct a new current query by replacing the corresponding subquery with the corresponding partial answer, and retrieving a set of ranked search results corresponding to the new current query; and presenting the set of ranked search results on a display device when the user satisfaction level exceeds the predetermined threshold. - View Dependent Claims (12, 13, 14, 15)
-
-
16. A computer-readable storage device having computer executable instructions stored therein for improving relevance of search results in response to a user query, said instructions comprising:
-
for receiving a current query from a user; evaluating a set of historical search logs for determining a user satisfaction level for a set of query results returned in response to the current query; iteratively reformulating the current query to construct a new current query, with such iterations repeating until the user satisfaction level exceeds a predetermined threshold, and wherein iteratively reformulating the current query comprises instructions for; evaluating the current query to identify one or more subqueries, each subquery comprised of one or more of the search terms of the current query, evaluating each subquery to select one or more structured data sets from among a plurality of structured data sets from which to retrieve a partial answer corresponding each subquery, for each partial answer retrieved from a selected structured data set, reformulating the current query to construct a new current query by replacing the corresponding subquery with the corresponding partial answer, and retrieving a set of ranked search results corresponding to the new current query; and presenting the set of ranked search results on a display device when the user satisfaction level exceeds the predetermined threshold. - View Dependent Claims (17, 18, 19, 20)
-
Specification