UNLEARNING TECHNIQUES FOR ADAPTIVE LANGUAGE MODELS IN TEXT ENTRY

• US 20160282956A1
• Filed: 10/01/2015
• Published: 09/29/2016
• Est. Priority Date: 03/24/2015
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Ouyang, Yu, Zhai, Shumin
• Granted Patent: US 9,703,394 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 2203/04808 Several contacts: gestures triggering a specific function, e.g. scrolling, zooming, right-click, when the user establishes several contacts with the surface simultaneously; e.g. using several fingers or a combination of fingers and pen
• CPC Class Codes: G06F 3/0237 using prediction or retrieval techniques
• CPC Class Codes: G06F 3/04883 for inputting data by handwriting, e.g. gesture or text
• CPC Class Codes: G06F 3/04886 by partitioning the display area of the touch-screen or the surface of the digitising tablet into independently controllable areas, e.g. virtual keyboards or menus
• CPC Class Codes: G06F 40/00 Handling natural language data speech analysis or synthesis, speech recognition G10L
• CPC Class Codes: G06F 40/12 Use of codes for handling textual entities
• CPC Class Codes: G06F 40/166 Editing, e.g. inserting or deleting
• CPC Class Codes: G06F 40/216 using statistical methods

Abstract

In some examples, a method includes outputting a graphical keyboard (120) for display and responsive to receiving an indication of a first input (124), determining a new character string that is not included in a language model. The method may include adding the new character string to the language model and associating a likelihood value with the new character string. The method may include, responsive to receiving an indication of a second input, predicting the new character string, and responsive to receiving an indication of a third input that rejects the new character string, decreasing the likelihood value associated with the new character string.

Claims

1. A method comprising:

outputting, by a computing device and for display at an output component operatively coupled to the computing device, a graphical keyboard;

responsive to receiving an indication of a first input detected at a location of a presence-sensitive input component operatively coupled to the computing device, determining, based at least in part on the indication of the first input, a new character string that is not included in a language model, wherein the location of the presence-sensitive input component corresponds to a location of the output component;

adding the new character string to the language model and associating a likelihood value with the new character string;

responsive to receiving an indication of a second input detected at a second location of the presence-sensitive input component, predicting the new character string; and

responsive to receiving an indication of a third input that rejects the new character string, decreasing the likelihood value associated with the new character string.

2. The method of claim 1, wherein associating the likelihood value with the new character string comprises storing the new character string in at least one n-gram of the language model with the likelihood value, such that the likelihood value corresponds to the at least one n-gram in the language model, the method further comprising:

responsive to receiving the indication of the third input that rejects the new character string;

determining a plurality of n-grams in the language model that include the new character string; and

decreasing a respective likelihood value that corresponds to each respective n-gram of the plurality of n-grams.

3. The method of claim 2, further comprising:

responsive to determining that a first likelihood value that corresponds a higher-order n-gram of the language model is greater than a second likelihood value that corresponds to a lower-order n-gram, modifying the language model to at least;

remove the higher-order n-gram, or

update the first likelihood value to be less than or equal to the second likelihood value,

wherein the higher-order n-gram includes more character strings than the lower-order n-gram.

4. The method of claim 1, wherein decreasing the likelihood value associated with the new character string further comprises decreasing the likelihood value associated with the new character string without removing the new character string from the language model.

5. The method of claim 1, further comprising:

determining a particular input context when predicting the new character string, wherein the particular input context comprises one or more previously inputted character strings that are outputted for display when predicting the new character string,

wherein decreasing the likelihood value associated with the new character string comprises decreasing the likelihood value associated with the new character string for only the particular input context.

6. The method of claim 1, further comprising:

determining a particular input context when predicting the new character string,

wherein decreasing the likelihood value associated with the new character string comprises decreasing the likelihood value associated with the new character string at a rate or magnitude that is based at least in part on the input context.

7. The method of claim 1, wherein a rate or magnitude of increasing the likelihood value that corresponds to the new character string in response to learning the new character string is different than a rate or magnitude of decreasing the likelihood value in response to unlearning the new character string.

8. The method of claim 1, wherein the third input that rejects the new character string comprises an input that:

deletes the new character string;

deletes a sequence of character strings that includes the new character string;

ignores the new character string;

cancels an auto-correction of the new character string;

or

corrects the new character string using a spell-checker.

9. The method of claim 1, wherein each of the first and second inputs comprise at least one of a discrete tap gesture or a portion of a continuous gesture.

10. A computing device comprising:

one or more computer processors;

an output component;

a presence-sensitive input component; and

a memory comprising instructions that when executed by the one or more computer processors cause the one or more computer processors to;

output, for display at the output component, a graphical keyboard;

responsive to receiving an indication of a first input detected at a location of the presence-sensitive input component, determine, based at least in part on the indication of the first input, a new character string that is not included in a language model, wherein the location of the presence-sensitive input component corresponds to a location of the output component;

add the new character string to the language model and associate a likelihood value with the new character string;

responsive to receiving an indication of a second input detected at a second location of the presence-sensitive input component, predict the new character string; and

responsive to receiving an indication of a third input that rejects the new character string, decrease the likelihood value associated with the new character string.

11. The computing device of claim 10, wherein the memory comprises instructions that when executed by the one or more computer processors cause the one or more computer processors to:

store the new character string in at least one n-gram of the language model with the likelihood value, such that the likelihood value corresponds to the at least one n-gram in the language model; and

responsive to receiving the indication of the third input that rejects the new character string;

determine a plurality of n-grams in the language model that include the new character string; and

decrease a respective likelihood value that corresponds to each respective n-gram of the plurality of n-grams.

12. The computing device of claim 11, wherein the memory comprises instructions that when executed by the one or more computer processors cause the one or more computer processors to:

responsive to determining that a first likelihood value that corresponds a higher-order n-gram of the language model is greater than a second likelihood value that corresponds to a lower-order n-gram, modify the language model to at least;

remove the higher-order n-gram, or

update the first likelihood value to be less than or equal to the second likelihood value,

wherein the higher-order n-gram includes more character strings than the lower-order n-gram.

13. The computing device of claim 10, wherein the memory comprises instructions that when executed by the one or more computer processors cause the one or more computer processors to:

decrease the likelihood value associated with the new character string without removing the new character string from the language model.

14. The computing device of claim 10, wherein the memory comprises instructions that when executed by the one or more computer processors cause the one or more computer processors to:

determine a particular input context when predicting the new character string, wherein the particular input context comprises one or more previously inputted character strings that are outputted for display when predicting the new character string; and

decrease the likelihood value associated with the new character string for only the particular input context.

15. The computing device of claim 10, wherein each of the first and second inputs comprise at least one of a discrete tap gesture or a portion of a continuous gesture.

16. A non-transitory computer-readable storage medium encoded with instructions that, when executed, cause at least one processor of a computing device to:

output, for display at an output component operatively coupled to the computing device, a graphical keyboard;

responsive to receiving an indication of a first input detected at a location of a presence-sensitive input component operatively coupled to the computing device, determine, based at least in part on the indication of the first input, a new character string that is not included in a language model, wherein the location of the presence-sensitive input component corresponds to a location of the output component;

add the new character string to the language model and associate a likelihood value with the new character string;

responsive to receiving an indication of a second input detected at a second location of the presence-sensitive input component, predict the new character string; and

responsive to receiving an indication of a third input that rejects the new character string, decrease the likelihood value associated with the new character string.

17. The non-transitory computer-readable storage medium of claim 16 encoded with instructions that, when executed, cause the at least one processor of the computing device to:

store the new character string in at least one n-gram of the language model with the likelihood value, such that the likelihood value corresponds to the at least one n-gram in the language model; and

responsive to receiving the indication of the third input that rejects the new character string;

determine a plurality of n-grams in the language model that include the new character string; and

decrease a respective likelihood value that corresponds to each respective n-gram of the plurality of n-grams.

18. The non-transitory computer-readable storage medium of claim 17 encoded with instructions that, when executed, cause the at least one processor of the computing device to:

responsive to determining that a first likelihood value that corresponds a higher-order n-gram of the language model is greater than a second likelihood value that corresponds to a lower-order n-gram, modify the language model to at least;

remove the higher-order n-gram, or

update the first likelihood value to be less than or equal to the second likelihood value,

wherein the higher-order n-gram includes more character strings than the lower-order n-gram.

19. The non-transitory computer-readable storage medium of claim 16 encoded with instructions that, when executed, cause the at least one processor of the computing device to:

decrease the likelihood value associated with the new character string without removing the new character string from the language model.

20. The non-transitory computer-readable storage medium of claim 16 encoded with instructions that, when executed, cause the at least one processor of the computing device to:

determine a particular input context when predicting the new character string, wherein the particular input context comprises one or more previously inputted character strings that are outputted for display when predicting the new character string, and

decrease the likelihood value associated with the new character string for only the particular input context.

21. The non-transitory computer-readable storage medium of claim 16, wherein each of the first and second inputs comprise at least one of a discrete tap gesture or a portion of a continuous gesture.