Touch keyboard using a trained model

1. A method comprising:

outputting, by a computing device, for display at a presence-sensitive display, a graphical user interface including a plurality of keys of a graphical keyboard;

executing a model trained using machine-learning to determine text from both sliding and single-touch gestures detected by the presence-sensitive display, the model including a spatial model and a language model, the spatial model including a distribution of touch points that corresponds to at least one respective key of the graphical keyboard;

responsive to receiving first gesture data from the presence-sensitive display, inputting the first gesture data into the model, the first gesture data indicative of one or more first letters selected by a first gesture type that comprises a sliding gesture or a single-touch gesture to select one or more first letter keys of the graphical keyboard, the sliding gesture comprising a continuous gesture over different locations of the graphical user interface, wherein alignment points along the continuous gesture correspond to different letter keys of the graphical keyboard;

responsive to receiving second gesture data from the presence-sensitive display, inputting the second gesture data into the model, the second gesture data indicative of one or more second letters selected by a second gesture type that comprises the sliding gesture or the single-touch gesture to select one or more second letter keys of the graphical keyboard, the second gesture type different than the first gesture type;

determining, based at least in part on both physical cost values from the spatial model and lexical cost values from the language model, one or more words predicted by the model based on the first and second letters, the physical cost values representing first likelihoods that the first gesture data and the second gesture data correspond to the first letters and the second letters, respectively, and modified by the lexical cost values from the language model, the lexical cost values representing second likelihoods that the first letters and the second letters are included in a word in a lexicon of the language model; and

outputting, by the computing device, for display in a text entry area of the graphical user interface, the one or more words predicted by the model.

2. The method of claim 1 , wherein the first gesture type is the sliding gesture and the second gesture type is the single-touch gesture.

3. The method of claim 1 , wherein the model is trained using statistical machine-learning.

4. The method of claim 1 , further comprising:

outputting, by the computing device, for display as a single prediction in a suggestion region of the graphical keyboard, the one or more words predicted by the model.

5. The method of claim 4 , further comprising:

while outputting the one or more words predicted by the model for display in the suggestion region of the graphical keyboard, outputting, for display, additional words predicted by the model as additional predictions in the suggestion region.

6. The method of claim 4 , wherein the suggestion region comprises up to three predictions including the one or more words as a first of the three predictions.

7. The method of claim 1 , wherein the model is trained using machine-learning to determine the one or more words by inferring spatial and linguistic characteristics of the first gesture data relative to spatial and linguistic characteristics of the second gesture data.

8. The method of claim 1 , the method further comprises:

determining a speed at which the first gesture data and the second gesture data are performed; and

modifying, based on the speed of the first gesture data and the second gesture data, one or more physical cost values of the spatial model,

wherein the model is trained using machine-learning to determine the one or more words by inferring the spatial characteristics of the first gesture data and the second gesture data relative to the one or more physical cost values of the spatial model.

9. A computing device comprising:

a presence-sensitive display; and

one or more processors configured to:

output, for display at the presence-sensitive display, a graphical user interface including a plurality of keys of a graphical keyboard;

execute a model trained using machine-learning to determine text from both sliding gestures and single-touch gestures detected by the presence-sensitive display, the model including a spatial model and a language model, the spatial model including a distribution of touch points that corresponds to at least one respective key of the graphical keyboard;

responsive to receiving first gesture data from the presence-sensitive display, input the first gesture data into the model, the first gesture data indicative of one or more first letters selected by a first gesture type that comprises a sliding gesture or a single-touch gesture to select one or more first letter keys of the graphical keyboard, the sliding gesture comprising a continuous gesture over different locations of the graphical user interface, wherein alignment points along the continuous gesture correspond to different letter keys of the graphical keyboard;

responsive to receiving second gesture data from the presence-sensitive display, input the second gesture data into the model, the second gesture data indicative of one or more second letters selected by a second gesture type that comprises the sliding gesture or the single-touch gesture to select one or more second letter keys of the graphical keyboard, the second gesture type different than the first gesture type;

determine, based at least in part on both physical cost values from the spatial model and lexical cost values from the language model, one or more words predicted by the model based on the first and second letters, the physical cost values representing first likelihoods that the first gesture data and the second gesture data correspond to the first letters and the second letters, respectively, and modified by the lexical cost values from the language model, the lexical cost values representing second likelihoods that the first letters and the second letters are included in a word in a lexicon of the language model; and

output, for display as a single prediction in a suggestion region of the graphical keyboard, the one or more words predicted by the model.

10. The computing device of claim 9 , wherein the first gesture type is the sliding gesture and the second gesture type is the single-touch gesture.

11. The computing device of claim 9 , wherein the model is trained using statistical machine-learning.

12. The computing device of claim 9 , wherein the one or more processors are further configured to:

output, for display in a text entry region of the graphical keyboard, the one or more words predicted by the model.

13. The computing device of claim 9 , wherein the one or more processors are further configured to:

while outputting the one or more words predicted by the model for display in the suggestion region of the graphical keyboard, output, for display, additional words predicted by the model as additional predictions in the suggestion region.

14. The computing device of claim 9 , wherein the suggestion region comprises up to three predictions including the one or more words as a first of the three predictions.

15. The computing device of claim 9 , wherein the model is trained using machine-learning to determine the one or more words by inferring spatial and linguistic characteristics of the first gesture data relative to spatial and linguistic characteristics of the second gesture data.

16. The computing device of claim 9 , wherein the one or more processors are further configured to:

determine a speed at which the first gesture data and the second gesture data are performed; and

modify, based on the speed of the first gesture data and the second gesture data, one or more physical cost values of the spatial model,

wherein the model is trained using machine-learning to determine the one or more words by inferring the spatial characteristics of the first gesture data and the second gesture data relative to the one or more physical cost values of the spatial model.

17. A non-transitory, non-signal computer-readable storage medium comprising instructions that, when executed by one or more processors, cause the one or more processors to:

output, for display at a presence-sensitive display, a graphical user interface including a plurality of keys of a graphical keyboard;

execute a model trained using machine-learning to determine text from both sliding gestures and single-touch gestures detected by the presence-sensitive display, the model including a spatial model and a language model, the spatial model including a distribution of touch points that corresponds to at least one respective key of the graphical keyboard;

responsive to receiving first gesture data from the presence-sensitive display, input the first gesture data into the model, the first gesture data indicative of one or more first letters selected by a first gesture type that comprises a sliding gesture or a single-touch gesture to select one or more first letter keys of the graphical keyboard, the sliding gesture comprising a continuous gesture over different locations of the graphical user interface, wherein alignment points along the continuous gesture correspond to different letter keys of the graphical keyboard;

responsive to receiving second gesture data from the presence-sensitive display, input the second gesture data into the model, the second gesture data indicative of one or more second letters selected by a second gesture type that comprises the sliding gesture or the single-touch gesture to select one or more second letter keys of the graphical keyboard, the second gesture type different than the first gesture type;

determine, based at least in part on both physical cost values from the spatial model and lexical cost values from the language model, one or more words predicted by the model based on the first and second letters, the physical cost values representing first likelihoods that the first gesture data and the second gesture data correspond to the first letters and the second letters, respectively, and modified by the lexical cost values from the language model, the lexical cost values representing second likelihoods that the first letters and the second letters are included in a word in a lexicon of the language model; and

output, for display as a single prediction in a suggestion region of the graphical keyboard, the one or more words predicted by the model.

18. The non-transitory, non-signal computer-readable storage medium of claim 17 , wherein the first gesture type is the sliding gesture and the second gesture type is the single-touch gesture.

19. The non-transitory, non-signal computer-readable storage medium of claim 17 , wherein the model is trained using statistical machine-learning.

20. The non-transitory, non-signal computer-readable storage medium of claim 17 comprising further instructions that, when executed by the one or more processors, cause the one or more processors to:

determine a speed at which the first gesture data and the second gesture data were performed; and

modify, based on the speed of the first gesture data and the second gesture data, one or more physical cost values of the spatial model,

wherein the model is trained using machine-learning to determine the one or more words by inferring the spatial characteristics of the first gesture data and the second gesture data relative to the one or more physical cost values of the spatial model.