Search result ranking using machine learning

1. A computer-implemented method for ordering search results in an online publishing system, the method comprising:

receiving, by a server device, a first input label specifying at least a first goal;

identifying, by the server device, a first set of data factor definitions based on the first input label;

compiling, by the server device, a first set of data factors using the first set of data factor definitions;

training, by the server device, a machine learning system using the first input label and the first set of data factors to create a first goal model for the first goal;

receiving, by the server device, a second input label specifying at least a second goal, the second goal being different from the first goal;

identifying, by the server device, a second set of data factor definitions based on the second input label, wherein the second set of data factor definitions is different than the first set of data factor definitions;

compiling, by the server device, a second set of data factors using the second set of data factor definitions;

training, by the server device, the machine learning system using the second input label and the second set of data factors to create a second goal model for the second goal;

receiving, by the server device, an electronic search query and a selection of the first input label or the second input label; and

ordering, by the server device, search results by applying either the first goal model or the second goal model to the electronic search query based on the selection of the first input label or the second input label respectively, wherein applying either the first goal model or the second goal model comprises factorizing the electronic search query into either the first set of data factors or the second set of data factors, respectively, and inputting a factorized electronic search query into either the first goal model or the second goal model based on the selection of the first input label or the second input label.

2. The method of claim 1 , wherein a data factor definition includes an identification of one or more data points that make up a corresponding data factor.

3. The method of claim 2 , wherein compiling a set of data factors includes retrieving the one or more data points and factorizing the one or more data points into a standard format corresponding to an input label.

4. The method of claim 3 , wherein factorizing the one or more data points includes aggregating a plurality of data points.

5. The method of claim 1 , wherein a subset of data factors in the first set of data factors each have a first impact score corresponding to the first input label and also each have a second impact score corresponding to the second input label, impact scores measuring the degree to which corresponding data factors influence an output of a corresponding goal model.

6. The method of claim 5 , wherein applying either the first goal model or the second goal model includes ranking either the first set of data factors or the second set of data factors using impact scores of respective data factors and using a group of either the first set of data factors or the second set of data factors based on the ranking and the selection of the first input label or the second input label.

7. The method of claim 1 , wherein the first goal comprises one of maximizing revenue or cross product marketing.

8. A machine readable medium that is not a transitory propagating signal, the machine readable medium including instructions that, when executed by a machine, cause the machine to perform operations for ordering search results in an online publishing system, the operations comprising:

receiving a first input label specifying at least a first goal;

identifying a first set of data factor definitions based on the first input label;

compiling a first set of data factors using the first set of data factor definitions;

training a machine learning system using the first input label and the first set of data factors to create a first goal model for the first goal;

receiving a second input label specifying at least a second goal, the second goal being different from the first goal;

identifying a second set of data factor definitions corresponding based on the second input label, wherein the second set of data factor definitions is different than the first set of data factor definitions;

compiling a second set of data factors using the second set of data factor definitions;

training the machine learning system using the second input label and the second set of data factors to create a second goal model for the second goal;

receiving, by the server device, an electronic search query and a selection of the first input label or the second input label; and

ordering search results by applying either the first goal model or the second goal model to the electronic search query based on the selection of the first input label or the second input label respectively, wherein applying either the first goal model or the second goal model comprises factorizing the electronic search query into either the first set of data factors or the second set of data factors, respectively, and inputting a factorized electronic search query into either the first goal model or the second goal model based on the selection of the first input label or the second input label.

9. The machine readable medium of claim 8 , wherein a data factor definition includes an identification of one or more data points that make up a corresponding data factor.

10. The machine readable medium of claim 9 , wherein compiling a set of data factors includes retrieving the one or more data points and factorizing the one or more data points into a standard format corresponding to an input label.

11. The machine readable medium of claim 10 , wherein factorizing the one or more data points includes aggregating a plurality of data points.

12. The machine readable medium of claim 8 , wherein a subset of data factors in the first set of data factors each have a first impact score corresponding to the first label and also each have a second impact score corresponding to the second input label, impact scores measuring the degree to which corresponding data factors influence an output of a corresponding goal model.

13. The machine readable medium of claim 12 , wherein applying either the first goal model or the second goal model includes ranking either the first set of data factors or the second set of data factors using impact scores of respective data factors and using a subset of either the first set of data factors or the second set of data factors based on the ranking and the selection of the first input label or the second input label.

14. A system for ordering search results in an online marketplace, the system, comprising:

execution hardware,

a machine learning module, performed by the execution hardware, configured to:

receive a first input label specifying at least a first goal;

identify a first set of data factor definitions based on the first input label;

compile a first set of data factors using the first set of data factor definitions;

receive a second input label specifying at least a second goal, the second goal being different from the first goal;

identify a second set of data factor definitions based on the second input label, wherein the second set of data factor definitions is different than the first set of data factor definitions; and

compile a second set of data factors using the second set of data factor definitions;

a training module, performed by the execution hardware, configured to:

train a machine learning system using the first input label and the first set of data factors to create a first goal model for the first goal; and

train the machine learning system using the second input label and the second set of data factors to create a second goal model for the second goal; and

a goal module, performed by the execution hardware, configured to receive an electronic search query and a selection of the first input label or the second input label and configured to order search results by applying either the first goal model or the second goal model to the electronic search query based on the selection of the first input label or the second input label respectively wherein applying either the first goal model or the second goal model comprises factorizing the electronic search query into either the first set of data factors or the second set of data factors, respectively, and inputting a factorized electronic search query into either the first goal model or the second goal model based on the selection of the first input label or the second input label.

15. The system of claim 14 , wherein a data factor definition includes an identification of one or more data points that make up a corresponding data factor.

16. The system of claim 15 , wherein to compile a set of data factors includes the machine learning module to retrieve the one or more data points and factorize the one or more data points into a standard format corresponding to an input label.

17. The system of claim 16 , wherein to factorize the one or more data points includes aggregating a plurality of data points.

18. The system of claim 14 , wherein a subset of data factors in the first set of data factors each have a first impact score corresponding to the first label and also each have a second impact score corresponding to the second input label, impact scores measuring the degree to which corresponding data factors influence a output of a corresponding goal model.

19. The system of claim 18 , wherein to apply either the first goal model or the second goal model includes the machine learning module to rank either the first set of data factors or the second set of data factors using impact scores of respective data factors and using a subset of either the first set of data factors or the second set of data factors based on the ranking and the selection of the first input label or the second input label.