Ranking search results using hierarchically organized machine learning based models

1. A method comprising:

storing a hierarchy of machine-learned models, wherein machine-learned models of the hierarchy are configured to receive as input a set of search results and generate as output scores for ranking the set of search results, the hierarchy of machine-learned models comprising a plurality of levels, levels of the hierarchy associated with respective sets of dimensions describing search queries, wherein machine-learned models at a level are associated with a respective n-tuple of values corresponding to the set of dimensions for that level;

training a particular machine-learned model using queries associated with a particular n-tuple of values;

adding the particular machine-learned model to the hierarchy of machine-learned models in a level corresponding to the n-tuple of values;

matching a new search query to the particular machine-learned model based on a respective set of dimensions of the search query; and

ranking search results of the new search query using the particular machine-learned model responsive to the dimensions of the new search query matching the particular machine-learned model.

2. The method of claim 1 , further comprising:

evaluating the performance of a set of search queries;

wherein training the particular machine-learned model is responsive to the evaluation.

3. The method of claim 2 , wherein evaluating the performance of the set of search queries comprises determining a score for a model corresponding to the set of search queries based on an aggregate measure of quality of a plurality of searches performed using the model, the measure of quality of a search determined using a rank of a search result accessed by a user from a set of ranked search results provided to the user.

4. The method of claim 1 , further comprising:

training the particular machine-learned model using previous search queries and a rank of a search result that a user accessed from each set of search results provided to the user in response to one of the previous search queries.

5. The method of claim 1 , wherein a first level of the hierarchy of machine-learned models comprises a base model, the base model associated with a first set of dimensions representing an empty set, the first level including a base model applicable to a plurality of entities and a plurality of organizations.

6. The method of claim 5 , wherein the hierarchy of machine-learned models comprises a second level associated with a second set of dimensions including an organization dimension, wherein the second level comprises a second set of models, each model for a specific organization.

7. The method of claim 6 , wherein the hierarchy of machine-learned models comprises a third level associated with a third set of dimensions including an entity type dimension, wherein the third level comprises a third set of models, each model for a specific entity type.

8. The method of claim 7 , wherein the hierarchy of machine-learned models comprises a fourth level associated with a fourth set of dimensions including the organization dimension and the entity type dimension, the fourth level comprising a fourth set of models, each model for a specific organization and a specific entity type.

9. A method for ranking search results in a multi-tenant system, the method comprising:

receiving a search query from a user;

identifying a plurality of dimensions of the search query;

receiving a plurality of search results from a search engine corresponding to the search query;

accessing a hierarchy of machine-learned models, wherein machine-learned models of the hierarchy are configured to receive as input a set of search results and generate as output scores for ranking the set of search results, the hierarchy of machine-learned models comprising a plurality of levels, levels of the hierarchy associated with respective sets of dimensions describing search queries, wherein machine-learned models at a level are associated with a respective n-tuple of values corresponding to the set of dimensions for that level;

identifying a machine-learned model that matches the dimensions of the search query; and

using the identified machine-learned model to rank the search results.

10. The method of claim 9 , wherein identifying the machine-learned model is responsive to an entity associated with the search query or an organization associated with the search query.

11. The method of claim 9 , wherein a first level of the hierarchy of machine-learned models comprises a base model, the base model associated with a first set of dimensions representing an empty set, the first level including a base model applicable to a plurality of entities and a plurality of organizations.

12. The method of claim 11 , wherein the hierarchy of machine-learned models comprises a second level associated with a second set of dimensions including an organization dimension, wherein the second level comprises a second set of models, each model for a specific organization.

13. The method of claim 12 , wherein the hierarchy of machine-learned models comprises a third level associated with a third set of dimensions including an entity type dimension, wherein the third level comprises a third set of models, each model for a specific entity type.

14. The method of claim 13 , wherein the hierarchy of machine-learned models comprises a fourth level associated with a fourth set of dimensions including the organization dimension and the entity type dimension, the fourth level comprising a fourth set of models, each model for a specific organization and a specific entity type.

15. A non-transitory computer readable storage medium, storing instructions for ranking search results in a multi-tenant system, the instructions for:

receiving a search query from a user;

identifying a plurality of dimensions of the search query;

receiving a plurality of search results from a search engine corresponding to the search query;

accessing a hierarchy of machine-learned models, wherein machine-learned models of the hierarchy are configured to receive as input a set of search results and generate as output scores for ranking the set of search results, the hierarchy of machine-learned models comprising a plurality of levels, levels of the hierarchy associated with respective sets of dimensions describing search queries, wherein machine-learned models at a level are associated with a respective n-tuple of values corresponding to the set of dimensions for that level;

identifying a machine-learned model that matches the dimensions of the search query; and

using the identified machine-learned model to rank the search results.

16. The non-transitory computer readable storage medium of claim 15 , wherein identifying the machine-learned model is responsive to an entity associated with the search query or an organization associated with the search query.

17. The non-transitory computer readable storage medium of claim 15 , wherein a first level of the hierarchy of machine-learned models comprises a base model, the base model associated with a first set of dimensions representing an empty set, the first level including a base model applicable to a plurality of entities and a plurality of organizations.

18. The non-transitory computer readable storage medium of claim 17 , wherein the hierarchy of machine-learned models comprises a second level associated with a second set of dimensions including an organization dimension, wherein the second level comprises a second set of models, each model for a specific organization.

19. The non-transitory computer readable storage medium of claim 18 , wherein the hierarchy of machine-learned models comprises a third level associated with a third set of dimensions including an entity type dimension, wherein the third level comprises a third set of models, each model for a specific entity type.

20. The non-transitory computer readable storage medium of claim 19 , wherein the hierarchy of machine-learned models comprises a fourth level associated with a fourth set of dimensions including the organization dimension and the entity type dimension, the fourth level comprising a fourth set of models, each model for a specific organization and a specific entity type.