Observation-based query interpretation model modification

Patent metadata

Field	Value
Publication number	US-9798801-B2
Application number	US-201414333128-A
Country	US
Kind code	B2
Filing date	Jul 16, 2014
Priority date	Jul 16, 2014
Publication date	Oct 24, 2017
Grant date	Oct 24, 2017

Abstract

The modifying of a natural language interpretation model for interpreting natural language queries. The system discovers modifications that one or more queriers made to one or more original query results of one or more natural language queries to generate one or more modified query results. The system then uses the discoveries to identify one or more changes to a natural language interpretation model that would result (given the same natural language queries) in one or more query results that more accurately reflect the one or more modified query results. The system the causes the natural language interpretation model to be modified with at least one of the one or more identified changes. Accordingly, over time, the natural language interpretation model may learn from observations of its own performance.

First claim

What is claimed is:

1. A computer program product comprising one or more computer-readable storage media having stored thereon computer-executable instructions that are executable by one or more processors of a computer system to cause the computer system to modify a natural language interpretation model for natural language queries, the computer-executable instructions including instructions that are executable to cause the computer system to perform at least the following:
receive a first natural language query from a querier;
in response to receiving the first query, generate a first set of query results;
send the first set of query results to the querier;
receive a second query from the querier, the second query comprising a modified version of the first query that has been modified by the querier;
in response to receiving the second query, generate a second set of query results;
identify, within the second query, one or more modifications that the querier made to the first query;
based on the one or more identified modifications that the querier made to the first query, identify one or more changes to a natural language interpretation model that would result in query results that more accurately reflect the second set of query results as compared to the first set of query results when receiving a query that is the same as the first query, at least one of the one or more identified changes being identified as only applying to a particular querier or a particular group of queriers, such that when modifications of the natural language interpretation model corresponding to the at least one of the one or more identified changes is made, the modifications will only be applied to the particular querier or the particular group of queriers, rather than to every querier that submits a query;
modify, at a point in time, the natural language interpretation model with at least one of the one or more identified changes when at least one of the one or more identified changes surpasses an evidence threshold, wherein modifications that apply only to the particular group of queriers include a higher evidence threshold than modifications that apply either to all queriers or only to the particular querier;
receive a third query that is the same as the first query after the point time; and
in response to receiving the third query, and based at least on the third query having been received from the particular querier or from the particular group of queriers, generate the second set of query results using the modified natural interpretation model.

2. The computer program product in accordance with claim 1 , at least one of the first set of query results comprising a table, and the second set of query results comprising a modified table.

3. The computer program product in accordance with claim 1 , wherein the third query is received from a second querier who is a member of the particular group of queriers.

4. The computer program product in accordance with claim 1 , wherein causing the natural language interpretation model to be modified comprises causing the natural language interpretation model to be modified only as applied to the particular querier or the particular group of queriers.

5. The computer program product in accordance with claim 1 , wherein for each of a plurality of possible changes to the natural language interpretation model, identifying one or more changes to the natural language interpretation model comprises comparing an evidence threshold against corresponding modifications to the first set of query results, the evidence threshold not being the same for each of the plurality of possible changes to the natural language interpretation model.

6. The computer program product in accordance with claim 5 , wherein for each of at least one of the plurality of possible changes to the natural language interpretation model, identifying one or more changes to the natural language interpretation model also comprises adjusting the evidence threshold for the corresponding possible change to the natural language interpretation model in accordance with a confidence of the natural language interpretation model on a model portion that would be changed by the corresponding change.

7. The computer system in accordance with claim 1 , at least one of the first set of query results comprising a table, and the second set of query results comprising a modified table.

8. A method, implemented at a computer system that includes one or more processors, for modifying a natural language interpretation model for natural language queries, comprising:
receiving a first natural language query from a querier;
in response to receiving the first query, generating a first set of query results;
sending the first set of query results to the querier;
receiving a second query from the querier, the second query comprising a modified version of the first query that has been modified by the querier;
in response to receiving the second query, generating a second set of query results;
identifying, within the second query, one or more modifications that the querier made to the first query;
based on the one or more identified modifications that the querier made to the first query, identifying one or more changes to a natural language interpretation model that would result in query results that more accurately reflect the second set of query results as compared to the first set of query results when receiving a query that is the same as the first query, at least one of the one or more identified changes being identified as only applying to a particular querier or a particular group of queriers, such that when modifications of the natural language interpretation model corresponding to the at least one of the one or more identified changes is made, the modifications will only be applied to the particular querier or the particular group of queriers rather than to every querier that submits a query;
modifying, at a point in time, the natural language interpretation model with at least one of the one or more identified changes when at least one of the one or more identified changes surpasses an evidence threshold, wherein modifications that apply only to the particular group of queriers include a higher evidence threshold than modifications that apply either to all queriers or only to the particular querier;
receiving a third query that is the same as the first query after the point in time; and
in response to receiving the third query, and based at least on the third query having been received from the particular querier or from the particular group of queriers, generating the second set of query results using the modified natural interpretation model.

9. The method in accordance with claim 8 , at least one of the first set of query results comprising a table, and the second set of query results comprising a modified table.

10. The method in accordance with claim 9 , a modification causing the table to be modified to the modified table comprising:
changing one or more rows that are displayed.

11. The method in accordance with claim 9 , a modification causing the table to be modified to the modified table comprising:
changing one or more columns that are displayed.

12. The method in accordance with claim 9 , a modification causing the table to be modified to the modified table comprising:
applying one or more filters to the table to form the modified table.

13. The method in accordance with claim 9 , a modification causing the table to be modified to the modified table comprising:
applying one or more constraints to the table to form the modified table.

14. The metho

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Rezaei Pedram Faghihi
• Hays Christopher A
• Netz Amir M
• Baumgartner Patrick J

Classifications

• G06F17/30864

  Physics · mapped topic

• G06F17/30684 Primary

  Physics · mapped topic

• G06F17/30696

  Physics · mapped topic

• G06F17/30401

  Physics · mapped topic

• G06F16/243 Primary

  Natural language query formulation · CPC title