Query-goal-mission structures

1. A method, comprising:

evaluating a set of queries to identify query information for queries within the set of queries;

evaluating the queries as query pairs utilizing a goal classifier to determine common goal probabilities for the query pairs based upon the query information;

grouping one or more query pairs associated with common goal probabilities exceeding a goal probability threshold into a plurality of goal clusters;

evaluating the plurality of goal clusters as goal cluster pairs utilizing a mission classifier to determine common mission probabilities for the goal cluster pairs, wherein the plurality of goal clusters comprises at least a first goal cluster and a second goal cluster;

grouping the first goal cluster and the second goal cluster, of a first goal cluster pair of the goal cluster pairs, into a first mission cluster based upon a first common mission probability, for the first goal cluster pair, exceeding a mission probability threshold, wherein the first mission cluster is associated with two or more query pairs;

generating a query-goal-mission structure for the set of queries based upon the plurality of goal clusters and the first mission cluster;

receiving a search query from a remote device;

evaluating the search query to identify a search query aspect;

responsive to the search query aspect corresponding to an aspect of the query-goal-mission structure, using the query-goal-mission structure to identify a query recommendation; and

transmitting the query recommendation to the remote device.

2. The method of claim 1 , comprising:

performing a search assistance task utilizing the query-goal-mission structure.

3. The method of claim 2 , the performing a search assistance task comprising at least one of:

identifying an event recommendation;

identifying content associated with a product or service;

expanding a query submitted to a search engine; or

ranking search results.

4. The method of claim 2 , the performing a search assistance task comprising identifying content associated with a product or service.

5. The method of claim 2 , the performing a search assistance task comprising ranking search results.

6. The method of claim 1 , the evaluating the queries as query pairs comprising:

evaluating the query information to determine features for the query pairs; and

comparing the features to determine the common goal probabilities for the query pairs.

7. The method of claim 6 , the features comprising at least one of:

a query-pair local feature, a query-pair global feature, a query term-pair global feature, or a desktop query term-pair feature.

8. The method of claim 7 , the query-pair local feature comprising a conxsim feature.

9. The method of claim 6 , the goal classifier comprising a linear model with akaike information criterion, and the evaluating the query information to determine features for the query pairs comprising:

evaluating the query information for the query pairs utilizing the goal classifier to determine an aspect for the query pairs; and

select a first feature, but not a second feature, based upon the aspect.

10. The method of claim 1 , the evaluating the queries as query pairs comprising:

identifying a first set of features for the query pairs; and

utilizing the first set of features to determine the common goal probabilities for the query pairs; and

the evaluating the plurality of goal clusters comprising:

identifying a second set of features for the goal cluster pairs; and

utilizing the second set of features to determine the common mission probabilities for the goal cluster pairs, where the first set of features is different than the second set of features.

11. The method of claim 1 , the query information being indicative of a data sparseness concern, and the method comprising:

identifying the set of queries from a mobile search log;

evaluating the query information associated with the plurality of goal clusters to identify data sparseness; and

utilizing a desktop query term pair global feature to determine a common mission probability.

12. The method of claim 1 , the evaluating the queries comprising:

evaluating the query information to identify a number of long tail queries; and

responsive to the number of long tail queries exceeding a long tail query threshold, utilizing a query term pair global feature, but not a query pair global feature, to determine the common goal probabilities from the query information.

13. The method of claim 1 , wherein the set of queries contains queries associated with a cross-device search session.

14. The method of claim 1 , comprising:

training the goal classifier and the mission classifier to generate machine-learned rules for query classification, the training comprising:

evaluating a search log to identify a training dataset;

labeling the training data set to generate a ground truth dataset;

extracting features from the training dataset to generate a list of features; and

training at least one of the goal classifier or the mission classifier on the ground truth dataset to generate machine-learned rules for grouping queries into goal clusters and goal clusters into mission clusters.

15. A system, comprising:

a clustering component configured to:

evaluate a query session to identify query information for queries within the query session;

evaluate the queries as query pairs utilizing a goal classifier to determine common goal probabilities for the query pairs based upon the query information;

group one or more query pairs associated with common goal probabilities exceeding a goal probability threshold into a plurality of goal clusters;

evaluate the plurality of goal cluster as goal cluster pairs utilizing a mission classifier to determine common mission probabilities for the goal cluster pairs, wherein the plurality of goal clusters comprises at least a first goal cluster and a second goal cluster;

group the first goal cluster and the second goal cluster, of a first goal cluster pair of the goal cluster pairs, into a first mission cluster based upon a first common mission probability, for the first goal cluster pair, exceeding a mission probability threshold, wherein the first mission cluster is associated with two or more query pairs;

generate a query-goal-mission structure for the set of queries based upon the plurality of goal clusters and the first mission cluster;

receive a search query from a remote device;

evaluate the search query to identify a search query aspect;

responsive to the search query aspect corresponding to an aspect of the query-goal-mission structure, use the query-goal-mission structure to identify a query recommendation; and

transmit the query recommendation to the remote device.

16. The system of claim 15 , the query session comprising a cross device query session.

17. The system of claim 15 , comprising:

a search assistance component configured to at least one of:

identify an event recommendation based upon the query-goal-mission structure;

identify content associated with a product or service based upon the query-goal-mission structure; or

rank search results based upon the query-goal-mission structure.

18. The system of claim 15 , the clustering component configured to:

evaluate the query information to determine features for the query pairs; and

compare the features to determine the common goal probabilities for the query pairs.

19. A non-transitory computer readable medium comprising computer executable instructions that when executed by a processor perform a method, comprising:

evaluating a set of queries to identify query information for queries within the set of queries;

evaluating the queries as query pairs utilizing a goal classifier to determine common goal probabilities for the query pairs based upon the query information;

grouping one or more query pairs associated with common goal probabilities exceeding a goal probability threshold into a plurality of goal clusters;

evaluating the plurality of goal clusters as goal cluster pairs utilizing a mission classifier to determine common mission probabilities for the goal cluster pairs;

grouping the goal cluster pairs into mission clusters based upon the common mission probabilities for the goal cluster pairs exceeding a mission probability threshold;

generating a query-goal-mission structure for the set of queries based upon the plurality of goal clusters and the mission clusters;

receiving a search query from a remote device;

evaluating the search query to identify a search query aspect;

responsive to the search query aspect corresponding to an aspect of the query-goal-mission structure, using the query-goal-mission structure to identify a query recommendation; and

transmitting the query recommendation to the remote device.

20. The non-transitory computer readable medium of claim 19 , the features comprising at least one of:

a query-pair local feature, a query-pair global feature, a query term-pair global feature, or a desktop query term-pair feature.