Adaptive ranking of navigation suggestions

1 . A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause a computing device to perform operations comprising:

receiving an input in a navigation text box, the navigation text box being configured to receive text corresponding to a resource locator for a resource;

obtaining candidate navigation suggestions from a retrieval source for the input, the retrieval source representing data related to a browsing history for a user profile;

obtaining respective signals for the candidate navigation suggestions;

obtaining respective probability scores for the candidate navigation suggestions by providing the input, the candidate navigation suggestions, and the respective signals to a navigation suggestion ranking model, the respective probability scores representing a prediction of a likelihood of selection;

providing at least some candidate navigation suggestions based on their respective probability scores as autocomplete suggestions for the input; and

providing at least one search suggestion with the at least some candidate navigation suggestions,

wherein at least one of the candidate navigation suggestions is presented ahead of the at least one search suggestion based on the respective probability scores.

2 . The non-transitory computer-readable medium of claim 1 , the operations further comprising:

obtaining user signals describing a user,

wherein obtaining the respective probability scores includes providing the user signals with the input, the candidate navigation suggestions, and the respective signals to the navigation suggestion ranking model.

3 . The non-transitory computer-readable medium of claim 2 , the user signals including:

a user cohort determined for the user,

a device type,

a quantity of bookmarks saved on a client device,

a size of a browser history, or

a search-navigation ratio, the search-navigation ratio representing, for a number of prior user selections of the autocomplete suggestions in the navigation text box, a number of search suggestions selected and a number of navigation suggestions selected.

4 . The non-transitory computer-readable medium of claim 1 , the operations further comprising:

receiving a selection of an autocomplete suggestion of the autocomplete suggestions;

recording a selection event, the selection event including the respective signals, the candidate navigation suggestions, and the autocomplete suggestion selected; and

training the navigation suggestion ranking model using the selection event as labeled training data.

5 . The non-transitory computer-readable medium of claim 4 , wherein recording the selection event includes recording whether the selection is of a search suggestion or a candidate navigation suggestion.

6 . The non-transitory computer-readable medium of claim 1 , wherein a quantity of the at least some candidate navigation suggestions provided is determined by the navigation suggestion ranking model.

7 . The non-transitory computer-readable medium of claim 1 , wherein obtaining the respective signals further comprises:

deduplicating the candidate navigation suggestions from the retrieval source; and

annotating at least some of the candidate navigation suggestions with additional signals.

8 . The non-transitory computer-readable medium of claim 1 , the respective signals including, for a candidate navigation suggestion of the candidate navigation suggestions, a relative visit ratio.

9 . The non-transitory computer-readable medium of claim 1 , wherein a number of candidate navigation suggestions obtained from the retrieval source is an order of magnitude more than a number of candidate navigation suggestions provided for a rules-based ranking of the candidate navigation suggestions.

10 . The non-transitory computer-readable medium of claim 1 , wherein the respective signals include a score measuring a degree of match between the input and a title of a candidate navigation suggestion of the candidate navigation suggestions or a degree of match between the input and a domain of the candidate navigation suggestion.

11 . The non-transitory computer-readable medium of claim 1 , wherein the navigation suggestion ranking model runs on a client device.

12 . A computing device comprising:

at least a processor; and

a memory storing instructions that, when executed by the processor, cause the computing device to perform operations including:

receiving an input in a navigation text box, the navigation text box being configured to receive text corresponding to a resource locator for a resource;

obtaining candidate navigation suggestions from a retrieval source for the input, the retrieval source representing data related to a browsing history for a user profile;

obtaining respective signals for the candidate navigation suggestions;

obtaining respective probability scores for the candidate navigation suggestions by providing the input, the candidate navigation suggestions, and the respective signals to a navigation suggestion ranking model, the respective probability scores representing a prediction of a likelihood of selection;

providing at least some candidate navigation suggestions based on their respective probability scores as autocomplete suggestions for the input; and

providing at least one search suggestion with the at least some candidate navigation suggestions,

wherein at least one of the candidate navigation suggestions is presented ahead of the at least one search suggestion based on the respective probability scores.

13 . The computing device of claim 12 , the operations further comprising:

obtaining user signals describing a user,

wherein obtaining the respective probability scores includes providing the user signals with the input, the candidate navigation suggestions, and the respective signals to the navigation suggestion ranking model.

14 . The computing device of claim 13 , the user signals including:

a user cohort determined for the user,

a device type,

a quantity of bookmarks saved on a client device,

a size of a browser history, or

a search-navigation ratio, the search-navigation ratio representing, for a number of prior user selections of the autocomplete suggestions in the navigation text box, a number of search suggestions selected and a number of navigation suggestions selected.

15 . The computing device of claim 12 , the operations further comprising:

receiving a selection of an autocomplete suggestion of the autocomplete suggestions;

recording a selection event, the selection event including the respective signals, the candidate navigation suggestions, and the autocomplete suggestion selected; and

training the navigation suggestion ranking model using the selection event as labeled training data.

16 . The computing device of claim 15 , wherein recording the selection event includes recording whether the selection is of a search suggestion or a candidate navigation suggestion.

17 . The computing device of claim 12 , wherein a quantity of the at least some candidate navigation suggestions provided is determined by the navigation suggestion ranking model.

18 . The computing device of claim 12 , wherein obtaining the respective signals further comprises:

deduplicating the candidate navigation suggestions from the retrieval source; and

annotating at least some of the candidate navigation suggestions with additional signals.

19 . The computing device of claim 12 , the respective signals including, for a candidate navigation suggestion of the candidate navigation suggestions, a relative visit ratio.

20 . The computing device of claim 12 , wherein a number of candidate navigation suggestions obtained from the retrieval source is an order of magnitude more than a number of candidate navigation suggestions provided for a rules-based ranking of the candidate navigation suggestions.