Ranking items using a unified model

1. A method comprising:

receiving a contact list from a client device used by a user via a network, the contact list including contact information identifying contacts of the user in an online system;

receiving information about ephemeral content associated with contacts in the contact list, the ephemeral content posted by contacts in the contact list to the online system;

receiving information about which contacts in the contact list are currently using the online system;

determining scores for contacts in the contact list by applying a machine-learned model to the received information about ephemeral content and the received information about which contacts are currently using the online system, the machine-learned model outputting a score representing a likelihood that the user will interact with a contact in the contact list upon seeing the contact presented in an activity unit shown by the client device;

ranking the contacts in the contact list based at least in part on the determined scores; and

sending the ranked contact list to the client device.

2. The method of claim 1 , wherein the machined-learned model is a unified model trained using a plurality of data sets, different data sets containing different types of data having different features.

3. The method of claim 1 , further comprising:

collecting a training dataset, the training dataset including examples describing observed user interactions with ephemeral content and contacts that are currently using the online system;

extracting features from the examples of the training dataset features, the features relevant to likelihoods that the user will interact with a contact upon being presented with an indication that the contact is currently using the online system and/or has associated ephemeral content; and

training the machine-learned model using the features of the training dataset.

4. The method of claim 1 , wherein the machine-learned model determines the score responsive to a feature set comprising an active now modifier indicating whether the contact is currently using the online system and an ephemeral content modifier indicating whether the contact has available ephemeral content.

5. The method of claim 1 , wherein the machine-learned model selectively boosts a score associated with the contact to promote a particular interaction.

6. The method of claim 1 , wherein the client device is adapted to present the ranked contact list to the user an activity unit, the activity unit using different icons to represent different types of available interactions with the contacts.

7. A non-transitory computer-readable storage medium storing computer program instructions executable by a processor to perform operations comprising:

receiving a contact list from a client device used by a user via a network, the contact list including contact information identifying contacts of the user in an online system;

receiving information about ephemeral content associated with contacts in the contact list, the ephemeral content posted by contacts in the contact list to the online system;

receiving information about which contacts in the contact list are currently using the online system;

determining scores for contacts in the contact list by applying a machine-learned model to the received information about ephemeral content and the received information about which contacts are currently using the online system, the machine-learned model outputting a score representing a likelihood that the user will interact with a contact in the contact list upon seeing the contact presented in an activity unit shown by the client device;

ranking the contacts in the contact list based at least in part on the determined scores; and

sending the ranked contact list to the client device.

8. The computer-readable storage medium of claim 7 , wherein the machined-learned model is a unified model trained using a plurality of data sets, different data sets containing different types of data having different features.

9. The computer-readable storage medium of claim 7 , further comprising:

collecting a training dataset, the training dataset including examples describing observed user interactions with ephemeral content and contacts that are currently using the online system;

extracting features from the examples of the training dataset features, the features relevant to likelihoods that the user will interact with a contact upon being presented with an indication that the contact is currently using the online system and/or has associated ephemeral content; and

training the machine-learned model using the features of the training dataset.

10. The computer-readable storage medium of claim 7 , wherein the machine-learned model determines the score responsive to a feature set comprising an active now modifier indicating whether the contact is currently using the online system and an ephemeral content modifier indicating whether the contact has available ephemeral content.

11. The computer-readable storage medium of claim 7 , wherein the machine-learned model selectively boosts a score associated with the contact to promote a particular interaction.

12. The computer-readable storage medium of claim 7 , wherein the client device is adapted to present the ranked contact list to the user an activity unit, the activity unit using different icons to represent different types of available interactions with the contacts.

13. A system comprising:

a computer processor for executing computer program instructions; and

a non-transitory computer-readable storage medium storing computer program instructions executable by the processor to perform operations comprising:

receiving a contact list from a client device used by a user via a network, the contact list including contact information identifying contacts of the user in an online system;

receiving information about ephemeral content associated with contacts in the contact list, the ephemeral content posted by contacts in the contact list to the online system;

receiving information about which contacts in the contact list are currently using the online system;

determining scores for contacts in the contact list by applying a machine-learned model to the received information about ephemeral content and the received information about which contacts are currently using the online system, the machine-learned model outputting a score representing a likelihood that the user will interact with a contact in the contact list upon seeing the contact presented in an activity unit shown by the client device;

ranking the contacts in the contact list based at least in part on the determined scores; and

sending the ranked contact list to the client device.

14. The system of claim 13 , wherein the machined-learned model is a unified model trained using a plurality of data sets, different data sets containing different types of data having different features.

15. The system of claim 13 , further comprising:

collecting a training dataset, the training dataset including examples describing observed user interactions with ephemeral content and contacts that are currently using the online system;

extracting features from the examples of the training dataset features, the features relevant to likelihoods that the user will interact with a contact upon being presented with an indication that the contact is currently using the online system and/or has associated ephemeral content; and

training the machine-learned model using the features of the training dataset.

16. The system of claim 13 , wherein the machine-learned model determines the score responsive to a feature set comprising an active now modifier indicating whether the contact is currently using the online system and an ephemeral content modifier indicating whether the contact has available ephemeral content.

17. The system of claim 13 , wherein the machine-learned model selectively boosts a score associated with the contact to promote a particular interaction.