Optimally ranking accommodation listings based on constraints

1. A method comprising:

receiving, by an accommodation management system, from a client device of a user, a travel activity goal indicating a plurality of activities, each activity of the plurality of activities corresponding to a respective geographic region of a plurality of geographic regions;

determining a corresponding date for each activity of the plurality of activities;

determining for each activity of the plurality of activities, by the accommodation management system, a plurality of candidate accommodation listings based on geographic coordinates corresponding to the respective geographic region and the corresponding date for the activity;

ranking the plurality of candidate accommodation listings for each activity based on a selected mode of accessing the activity and one or more scores for each accommodation listing indicating a likelihood that the user would book the accommodation listing, wherein the one or more scores are generated by a machine learned model configured to output the one or more scores based on at least one of data describing the candidate accommodation listings, selected or inferred user preferences, or travel activity goal constraints; and

causing for display, on the client device, a recommendation for one or more accommodation listings for each activity of the plurality of activities from the candidate accommodation listings based on the ranking, wherein the recommendation for the one or more accommodation listings are grouped by respective activity.

2. The method of claim 1 , wherein the corresponding date for a first activity of the plurality of activities is a date range and further comprising:

determining that a single accommodation is not available for the date range of the first activity; and

creating a group of more than one candidate accommodation that satisfies the date range for the first activity.

3. The method of claim 1 , wherein determining for each activity of the plurality of activities, a plurality of candidate accommodation listings is further based on a set of data describing dates on which the user of the client device is available.

4. The method of claim 1 , wherein each candidate accommodation listing of the plurality of candidate accommodation listings is located at a geographic coordinate in a respective geographic region.

5. The method of claim 1 , wherein each candidate accommodation listing in the plurality of candidate accommodation listings is available during a time range corresponding to the date for the activity.

6. The method of claim 1 , wherein ranking the plurality candidate accommodation listings is further based on the user's accommodation preferences based on prior booking history in the accommodation management system.

7. The method of claim 6 , wherein ranking the plurality of candidate accommodation listings is further based on activity data from third-party systems describing the user's travel habit.

8. The method of claim 1 , wherein causing for display, on the client device, further comprises adding one or more available modes for accessing the activity corresponding to the candidate accommodation listing to the recommendation.

9. The method of claim 8 , wherein ranking the plurality of candidate accommodation listings further comprises:

displaying, by the accommodation management system, an interface on the client device, the interface including an interactable object for selecting from a plurality of preferred modes for accessing the activity;

receiving, by the accommodation management system, a selection of a preferred mode of accessing the activity by a user of the client device through an interaction with the interactable object; and

wherein the selection of the preferred mode of accessing the activity is the selected mode of accessing the activity.

10. The method of claim 1 , further comprising:

receiving, by the accommodation management system, text input, the text input including a description of a first activity of the plurality of activities;

inputting the text input into a machine learned model configured to determine a set of constraints;

receiving one or more constraints from the plurality of constraints as output from the machine learned model; and

before ranking the plurality of candidate accommodation listings for each activity, filtering the plurality of candidate accommodation listings by applying the one or more constraints output by the machine learning model.

11. The method of claim 10 , wherein receiving one or more constraints from the plurality of constraints as output from the machine learned model further comprises:

determining one or more words corresponding to one or more constraint types;

determining, by the accommodation management system, a set of data describing interactions by the user of the client device with the accommodation management system;

for each constraint category:

comparing the one or more words corresponding to the constraint type to the set of data; and

determining, based on the comparison, a constraint for the constraint type;

outputting the determined constraints for the one or more constraint categories.

12. The method of claim 1 , further comprising:

storing, by the accommodation management system, the recommendation with a plurality of recommendations corresponding to the user of the client device, each recommendation of the plurality of recommendations corresponding to one of a plurality of activity identifiers;

ranking, by the accommodation management system, the plurality of recommendations; and

sending for display, to the client device, a second recommendation for a second accommodation listing from the plurality of recommendations based on the ranking of the plurality of recommendations.

13. The method of claim 1 , further comprising:

receiving, by the accommodation management system, an update to an availability of a candidate accommodation listing from the plurality of candidate accommodation listings;

reranking the plurality of candidate accommodation listings based on the update; and

sending for display, to the client device, a second recommendation for a second accommodation listing based on the update.

14. The method of claim 1 , wherein the plurality of activities are listed in an order of the corresponding date for each activity.

15. The method of claim 14 , wherein the grouped recommendation for one or more accommodation listings are displayed with each respective activity listed in the order of the corresponding date for each activity.

16. The method of claim 14 , wherein each listed activity in the plurality of activities comprises an expandable drop-down menu for a respective grouped recommendation of one or more accommodation listings.

17. A system comprising:

a non-transitory computer-readable medium comprising instructions encoded thereon; and

one or more processors configured to execute the instructions, the instructions, when executed, causing the one or more processors to perform operations comprising:

receiving, from a client device of a user, a travel activity goal indicating a plurality of activities, each activity of the plurality of activities corresponding to a respective geographic region of a plurality of geographic regions;

determining a corresponding date for each activity of the plurality of activities;

determining for each activity of the plurality of activities, a plurality of candidate accommodation listings based on geographic coordinates corresponding to the respective geographic region and the corresponding date for the activity;

ranking the plurality of candidate accommodation listings for each activity based on a selected mode of accessing the activity and one or more scores for each accommodation listing indicating a likelihood that the user would book the accommodation listing, wherein the one or more scores are generated by a machine learned model configured to output the one or more scores based on at least one of data describing the candidate accommodation listings, selected or inferred user preferences, or travel activity goal constraints; and

causing for display, on the client device, a recommendation for one or more accommodation listings for each activity of the plurality of activities from the candidate accommodation listings based on the ranking, wherein the recommendation for the one or more accommodation listings are grouped by respective activity.

18. A non-transitory machine readable medium configured to store instructions, the instructions when executed by a processor cause the processor to perform operations, the instructions comprising instructions to perform operations comprising:

receiving, from a client device of a user, a travel activity goal indicating a plurality of activities, each activity of the plurality of activities corresponding to a respective geographic region of a plurality of geographic regions;

determining a corresponding date for each activity of the plurality of activities;

determining for each activity of the plurality of activities, a plurality of candidate accommodation listings based on geographic coordinates corresponding to the respective geographic region and the corresponding date for the activity;

ranking the plurality of candidate accommodation listings for each activity based on a selected mode of accessing the activity and one or more scores for each accommodation listing indicating a likelihood that the user would book the accommodation listing, wherein the one or more scores are generated by a machine learned model configured to output the one or more scores based on at least one of data describing the candidate accommodation listings, selected or inferred user preferences, or travel activity goal constraints; and

causing for display, on the client device, a recommendation for one or more accommodation listings for each activity of the plurality of activities from the candidate accommodation listings based on the ranking, wherein the recommendation for the one or more accommodation listings are grouped by respective activity.