Touch keyboard using language and spatial models

1. A method comprising:

outputting, by a computing device and for display, a graphical keyboard comprising a plurality of keys;

receiving, by the computing device, a plurality of indications of input, each respective indication of input from the plurality of indications of input corresponding to a respective location of the graphical keyboard; and

for each respective indication of input from the plurality of indications of input, incrementally:

determining, by the computing device and based at least in part on both physical cost values from a spatial model and lexical cost values from a language model, at least one predicted current word based on a set of characters that correspond to the plurality of indications of input, wherein the spatial model comprises at least one respective distribution of touch points that corresponds to at least one respective key of the graphical keyboard, and wherein the at least one predicted current word is determined based on a physical cost value from the spatial model that is modified by a lexical cost value from the language model, the physical cost value representing a first likelihood that the plurality of indications of input correspond to the set of characters, and the lexical cost value representing a second likelihood that the set of characters are included in any word in a lexicon of the language model;

determining, by the computing device and based at least in part on the at least one predicted current word, at least one predicted next word that follows the at least one predicted current word; and

outputting, by the computing device and for display, the at least one predicted current word and the at least one predicted next word as a soft commit word in a text entry area of a graphical user interface by at least:

outputting, for display in a first visual style, characters of the at least one predicted current word and the at least one predicted next word that are from the set of characters that correspond to the plurality of indications of input; and

outputting, for display in a second visual style that is different than the first visual style, characters of the at least one predicted current word and the at least one predicted next word that are not from the set of characters that correspond to the plurality of indications of input.

2. The method of claim 1 , further comprising:

for each respective indication of input from the plurality of indications of input, responsive to outputting at least the at least one predicted current word or the at least one predicted next word, incrementally modifying, by the computing device and based at least in part on a location of the key relative to a location of the respective indication of input, the spatial model.

3. The method of claim 1 , wherein determining the at least one predicted current word comprises performing, based at least in part on both the spatial model and the language model, error correction and word completion for each respective indication of input to determine the at least one predicted current word.

4. The method of claim 1 , wherein the set of characters is a portion of a complete word.

5. The method of claim 1 , wherein the language model comprises a lexicon that includes context information.

6. The method of claim 1 , wherein the language model comprises a word-level N-gram.

7. A computing device comprising:

at least one processor; and

memory configured to store instructions that, when executed, cause the at least one processor to:

output, for display, a graphical keyboard comprising a plurality of keys;

receive a plurality of indications of input, each respective indication of input from the plurality of indications of input corresponding to a respective location of the graphical keyboard; and

for each respective indication of input from the plurality of indications of input, incrementally:

determine, based at least in part on both physical cost values from a spatial model, and lexical cost values from a language model at least one predicted current word based on a set of characters that correspond to the plurality of indications of input, wherein the spatial model comprises at least one respective distribution of touch points that corresponds to at least one respective key of the graphical keyboard, and wherein the at least one predicted current word is determined based on a physical cost value from the spatial model that is modified by a lexical cost value from the language model, the physical cost value representing a first likelihood that the plurality of indications of input correspond to the set of characters, and the lexical cost value representing a second likelihood that the set of characters are included in any word in a lexicon of the language model;

determine, based at least in part on the at least one predicted current word that is the previous word, at least one predicted next word that follows the at least one predicted current word; and

output, for display, the at least one predicted current word and the at least one predicted next word as a soft commit word in a text entry area of a graphical user interface by at least:

outputting, for display in a first visual style, characters of the at least one predicted current word and the at least one predicted next word that are from the set of characters that correspond to the plurality of indications of input; and

outputting, for display in a second visual style that is different than the first visual style, characters of the at least one predicted current word and the at least one predicted next word that are not from the set of characters that correspond to the plurality of indications of input.

8. The computing device of claim 7 , wherein the instructions, when executed, further cause the at least one processor to determine a plurality of sets of one or more predicted current words, each respective set of one or more predicted current words from the plurality of sets of one or more predicted current words being based on a respective set of characters that correspond to the plurality of indications of input.

9. The computing device of claim 7 , wherein the instructions, when executed, further cause the at least one processor to perform error correction and word completion for each respective indication of input to determine the at least one predicted current word.

10. The computing device of claim 7 , wherein the instructions, when executed, further cause the at least one processor to:

determine, based at least in part on both a lexicon and a word-level N-gram, a frequency value associated with the set of characters, wherein the frequency value represents a frequency of occurrence of the set of characters in the lexicon; and

determine, based at least in part on the frequency value, the at least one predicted current word.

11. The computing device of claim 7 , wherein the lexicon comprises a listing of words and associated context information.

12. The computing device of claim 7 , wherein the set of characters is a portion of a complete word.

13. A non-transitory, non-signal computer-readable medium encoded with instructions that, when executed, cause at least one processor to:

output, for display, a graphical keyboard comprising a plurality of keys;

receive a plurality of indications of input, each respective indication of input from the plurality of indications of input corresponding to a respective location of an input device; and

for each respective indication of input from the plurality of indications of input, incrementally:

determine, based at least in part on both physical cost values from a spatial model, and lexical cost values from a language model, at least one predicted current word based on a set of characters that correspond to the plurality of indications of input, wherein the spatial model comprises at least one respective distribution of touch points that corresponds to at least one respective key of the graphical keyboard, and wherein the at least one predicted current word is determined based on a physical cost value from the spatial model that is modified by a lexical cost value from the language model, the physical cost value representing a first likelihood that the plurality of indications of input correspond to the set of characters, and the lexical cost value representing a second likelihood that the set of characters are included in any word in a lexicon of the language model;

determining, based at least in part on the at least one predicted current word, at least one predicted next word that follows the at least one predicted current word; and

output, for display, the at least one predicted current word and the at least one predicted next word as a soft commit word in a text entry area of a graphical user interface by at least:

outputting, for display in a first visual style, characters of the at least one predicted current word and the at least one predicted next word that are from the set of characters that correspond to the plurality of indications of input; and

outputting, for display in a second visual style that is different than the first visual style, characters of the at least one predicted current word and the at least one predicted next word that are not from the set of characters that correspond to the plurality of indications of input.

14. The non-transitory, non-signal computer-readable medium of claim 13 , further encoded with instructions that, when executed, cause the at least one processor to:

for each respective indication of input from the plurality of indications of input, responsive to outputting at least the at least one predicted current word or the at least one predicted next word, incrementally modify, by the computing device and based at least in part on a location of the key relative to a location of the respective indication of input, the spatial model.

15. The non-transitory, non-signal computer-readable medium of claim 13 , wherein the instructions that cause the at least one processor to determine the at least one predicted current word comprise instructions that, when executed, cause the at least one processor to perform, based at least in part on both the spatial model and the language model, error correction and word completion for each respective indication of input to determine the at least one predicted current word.

16. The non-transitory, non-signal computer-readable medium of claim 13 , wherein the at least one respective distribution of touch points comprises at least one respective two-dimensional Gaussian distribution.

17. The non-transitory, non-signal computer-readable medium of claim 13 , wherein the set of characters is a portion of a complete word.