SEARCH QUERIES WITH SHIFTING INTENT

• US 20100257164A1
• Filed: 04/07/2009
• Published: 10/07/2010
• Est. Priority Date: 04/07/2009
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Halverson, Alan Dale, Syed, Umar Ali, Slivkins, Aleksandrs, Mishra, Nina, Kenthapadi, Krishnaram
• Granted Patent: US 8,219,539 B2
• Time in Patent Office: Days
• US Class Current: 707/725
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques

Abstract

Techniques and systems are disclosed for returning temporally-aware results from an Internet-based search query. To determine if a query is temporally-based one or more query features are collected and input into a trained classifier, yielding a temporal classification for the query. Further, if a query is classified as temporal, the query results are shifted by determining an alternate set of results for the query, and returning one or more alternate results to one or more users. Based on user interactions with the one or more alternate results, the classifier can be updated, for example, by changing the query to a non-temporal query if the user interactions identify it as such.

Claims

1. A method for returning temporally-aware results from an Internet-based search query, comprising:

determining if a query is temporally-based, comprising;

collecting one or more features of the query; and

classifying whether the query is temporally-based by inputting the one or more features into a trained classifier;

shifting results for the query if the query is classified as temporally-based, comprising;

determining an alternate set of results for the query; and

returning one or more alternate results from the set of alternate results to one or more users; and

updating the classification of a query based on the one or more users'"'"' interactions with the one or more alternate results.

2. The method of claim 1, collecting one or more features of the query comprising one or more of:

determining how a user interacts with a first set of results returned for the query;

determining characteristics of the query; and

determining statistical-based characteristics of one or more of;

burstiness of user interactions with a first set of results returned for the query and characteristics of the query;

geographical location of the query;

age of document returned by query;

age of document in returned results selected by user;

a desired number of recent queries that are similar to the query; and

a desired number of recent news queries that are similar to the query.

3. The method of claim 1, determining if a query is temporally-based, comprising:

representing the one or more features of the query as a vector of coordinates in a feature space; and

labeling the vector of coordinates in a feature space for the query as temporally-based.

4. The method of claim 3, classifying whether the query is temporally-based comprising:

classifying the query as non-temporally based if the vector of coordinates for the query is within a non-temporally based query region of the feature space; and

classifying the query as temporally based if the vector of coordinates for the query is within a temporally-based query region of the feature space.

5. The method of claim 3, updating the classification of a query comprising:

changing the label of the vector of coordinates in a feature space for the query to non-temporally based if one or more users'"'"' interactions with the one or more alternate results indicate that the query is non-temporally based; and

updating a margin separating a temporally-based query region and a non-temporally based query region in the feature space.

6. The method of claim 2, determining how a user interacts with a first set of results returned for the query comprising determining one or more of:

a click-through rate comprising a rate that a user interacts with a returned result;

an abandonment rate comprising a rate that a query is abandoned after results are returned;

a reformulation rate comprising a rate at which a user does not interact with returned results and subsequently initiates an alternate query; and

a search engine switch rate comprising a rate that a user switches to an alternate search engine after results are returned.

7. The method of claim 2, determining characteristics of the query comprising determining one of:

a frequency rate comprising a rate that the query is posed;

a mention rate comprising a rate that the query is described in reference articles; and

an encyclopedia change comprising whether an article that comprises the query in an online encyclopedia has been recently updated.

8. The method of claim 1, determining an alternate set of results for the query comprising using an Internet-based search engine to identify a second set of results for the query having different characteristics than a first set of results identified by the search engine.

9. The method of claim 1, returning one or more alternate results from the set of alternate results to one or more users comprising returning an alternate result from a second set of results to a first set of one or more users for the respective alternate results in the second set of results, using a second set of one or more users for respective alternate results in the second set of alternate results, a desired number of times.

10. The method of claim 9, updating the classification of a query based on the one or more users'"'"' interactions with the one or more alternate results comprising labeling the query as non-temporal if a distribution of a click-through rate for the second set of returned results is unchanged from the first set of returned results within a desired threshold margin.

11. The method of claim 8, comprising changing the alternate set of results to meet one or more users'"'"' desired set of results based on one or more users'"'"' interactions with the alternate set of results.

12. The method of claim 1, updating the classification of a query based on the one or more users'"'"' interactions with the one or more alternate results comprising comparing the user interactions with the first set of results to the user interactions with one or more alternate results from the second set of results.

13. The method of claim 12, updating the classification of the query comprising labeling the query as non-temporal if the comparison of the user interactions indicates that the interactions are unchanged within a desired threshold margin.

14. The method of claim 11, updating the classification of the query comprising labeling the query as non-temporal if the alternate set of results is changed to a set of one or more results that is similar to the first set of one or more results returned by the search engine within a desired threshold margin.

15. A system for returning temporally-aware results from an Internet-based search query, comprising:

a feature collector configured to collect features of an Internet-based search query;

a query classifier operably coupled to the feature collector and configured to classify whether a query is temporal based on features provided by the feature collector;

a query results adapter configured to provide alternate results for the query if the query classifier determines that he query is temporal, comprising;

an alternate results determination component configured to use a search engine to determine a second set of results having different characteristics than a first set of results provided by the search engine; and

an alternate results provider configured to provide alternate results for one or more users; and

a query classifier updating component configured to update the query classifier to incorporate results from one or more users'"'"' interactions with the alternate results for a query.

16. The system of claim 15, comprising:

a feature space function component operably coupled to the query classifier configured to;

compute a feature space function for a vector of coordinates for features of a query; and

label the vector of coordinates for the query according to its computed result'"'"'s position in the feature space; and

a feature space updater configured to;

update the label of the computed query based on results from the query classifier updating component; and

update the feature space function so that function computation results for a temporally-based query region and function computation results for a non-temporally based query region in the feature space are separated by a margin.

17. The system of claim 16, the margin comprising:

a border area in feature space describing one of;

a space of temporal queries separated into half space by a hyper-plane;

a space of temporal queries separated by a convex hull; and

a space of temporal queries separated by a hyper-rectangle.

18. The system of claim 15, the query results adapter comprising an adapted results selector configured to combine a function of a click-through rate on returned alternate results with a function of a number of times corresponding returned alternate results have been shown to users to provide a set of one or more desired alternate results.

19. The system of claim 18, comprising a query rechecker configured to determine whether the query is temporal based on user interaction with the set of one or more alternate results.

20. A method for returning temporally-aware results from an Internet-based search query, comprising:

determining if a query is temporally-based, comprising;

collecting one or more features of the query, comprising;

determining how a user interacts with a first set of results returned for the query;

determining characteristics of the query; and

determining statistical-based characteristics of one or more of;

the user interactions with a first set of results returned for the query; and

the characteristics of the query; and

classifying whether the query is temporally-based, comprising;

inputting the one or more features into a trained classifier;

classifying the query as non-temporally based if the vector of coordinates for the query is within a non-temporally based query region of the feature space; and

classifying the query as temporally based if the vector of coordinates for the query is within a temporally-based query region of the feature space;

shifting results of the query if the query is classified as temporally-based, comprising;

determining an alternate set of results for the query using an Internet-based search engine to identify a second set of results for the query having different characteristics than a first set of results identified by the search engine; and

returning one or more alternate results from the set of alternate results to one or more users comprising returning an alternate result from a second set of results to a set of one or more users for the respective alternate results in the second set of results, using a different set of one or more users for respective alternate results in the second set of alternate results, a desired number of times; and

updating the classification of a query based on the one or more users'"'"' interactions with the one or more alternate results, comprising;

calculating the vector of coordinates for a query to feature space;

labeling the plotted query as temporally-based;

changing the label of the vector of coordinates in a feature space for the query to non-temporally based if one or more users'"'"' interactions with the one or more alternate results indicate that the query is non-temporally based; and

updating a margin separating a temporally-based query region and a non-temporally based query region in the feature space.