System and method for inputting text into electronic devices

• US 10,402,493 B2
• Filed: 07/18/2016
• Issued: 09/03/2019
• Est. Priority Date: 03/30/2009
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Touchtype Limited (Microsoft Corporation)
• Inventors: Spencer, Stephen Thomas, Creed, Paidi, Medlock, Benjamin William, Orr, Douglas Alexander Harper
• Primary Examiner(s): He, Jialong
• Application Number:
  US15/213,168
• Publication Number:
  US 20160328377A1
• Time in Patent Office: 1,142 Days
• CPC Class Codes: G06F 3/0237 using prediction or retrieval techniques
• 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 40/274 Converting codes to words; Guess-ahead of partial word inputs
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models

Abstract

Systems comprising a user interface configured to receive text input by a user and a text prediction engine configured to receive the input text and generate text predictions. The text prediction engine may comprise a general language model and a context-specific language model. The text prediction engine is configured to generate text predictions from the general language model and the context-specific language model and combine the text predictions. The text prediction engine may comprise first and second language models and a first context-specific weighting factor associated with the first language model. The text prediction engine is configured to generate text predictions using the first and second language models, generate weighted probabilities of the text predictions from the first language model using the first context-specific weighting factor; and generate final text predictions from the weighted predictions generated from the first language model and the predictions generated by the second language model.

Claims

1. A text prediction system comprising:

a processor; and

memory storing instructions that, when executed by the processor, configure the system to;

receive text input during a period of time;

identify a first time context corresponding to the text inputted during the period of time;

train a context-specific language model, associated with the first time context based at least in part on the text inputted during the period of time;

identify a second time context associated with text received during a second period of time, the second time context indicating that the second period of time includes a common time context with the period of time;

using the context-specific language model associated with the first time context and a general language model, concurrently generate a plurality of text predictions based at least in part on the text received during the second period of time, each of the plurality of text predictions comprising a term and an associated probability value; and

output concurrently at least one of the plurality of text predictions including at least one text prediction generated by the context-specific language model for the common time context and at least one text prediction generated by the general language model, wherein the at least one text prediction generated by the general language model is output based on a combined probability corresponding to the at least one text prediction generated by the general language, the combined probability including a context-specific weighting term when the at least one text prediction generated by the general language model is found in the context-specific language model.

2. The text prediction system of claim 1, wherein the context-specific language model is a time specific dynamic language model.

3. The text prediction system of claim 1, wherein the period of time comprises at least one of:

a day, a month, a year, or a weekend.

4. The text prediction system of claim 1, wherein the instructions further configure the system to:

generate at least one second text prediction during a third period of time, each second text prediction comprising a second term and a second associated probability value, based on the received text input, using the general language model; and

output the at least one second text prediction.

5. The text prediction system of claim 4, wherein the at least one text prediction or the at least one second text prediction comprises a word candidate, wherein when a word candidate generated using the general language model matches a word candidate generated using the context-specific language model, the system is configured to apply a weighting factor to a probability of the word candidate generated using the general language model, and wherein the weighting factor is based on a probability of that word generated using the context-specific language model.

6. The text prediction system of claim 4, wherein the at least one text prediction or the at least one second text prediction comprises a word candidate, and wherein when a word candidate generated using the general language model matches a word candidate generated using the context-specific language model, the system is configured to calculate a linear combination of the associated probability value generated by the general language model and the associated probability value generated by the context-specific language model for that matching word.

7. The text prediction system of claim 6, wherein the processor calculates a weighted linear combination, wherein the weighting factor is based on a ratio of the associated probability value of that word in the context-specific language model to the second associated probability value of that word in the general language model.

8. A text prediction system comprising:

a processor; and

memory storing instructions that, when executed by the processor, configure the text prediction system to;

receive text input during a period of time;

identify a first recipient context corresponding to a portion of the text inputted during the period of time;

train a plurality of recipient-specific language models based at least in part on the text inputted during the period of time, wherein each recipient-specific language model is trained on text sent to an associated recipient;

identify the recipient associated with the recipient context from text received during a second period of time;

using one of the plurality of recipient-specific language models associated with the first recipient context, and a general language model, concurrently generate a plurality of text predictions comprising a term and an associated probability value, based at least in part on the text received during the second period of time;

output concurrently at least one of the plurality of text predictions including at least one text prediction generated by the one of the plurality of recipient-specific language models for the first recipient context and at least one text prediction generated by the general language model, wherein the at least one text prediction generated by the general language model is output based on a combined probability corresponding to the at least one text prediction generated by the general language, the combined probability including a recipient-specific weighting term when the at least one text prediction generated by the general language model is found in one of the plurality of recipient-specific language models.

9. The text prediction system of claim 8, wherein the one of the plurality of recipient-specific language models is a dynamic language model.

10. The text prediction system of claim 8, wherein the one of the plurality of recipient-specific language models further includes a MessageThread-specific language model.

11. A method for predicting text, comprising:

accessing a plurality of recipient-specific language models trained on text previously sent to recipients associated with respective recipient-specific language models;

receiving text input;

identify a recipient context corresponding to the text input;

identify a recipient associated with the recipient context from the received text input;

using one of the plurality of recipient-specific language models associated with the recipient context and a general language model, concurrently generating a plurality of text predictions, each of the plurality of text predictions comprising a term and an associated probability value, based at least in part on the received text input; and

outputting concurrently at least one of the plurality of text predictions including a first text prediction generated by the one of the plurality of recipient-specific language models for the recipient context and a second text prediction generated by the general language model, wherein the at least one text prediction generated by the general language model is output based on a combined probability corresponding to the at least one text prediction generated by the general language, the combined probability including a recipient-specific weighting term when the at least one text prediction generated by the general language model is found in one of the plurality of recipient-specific language models.

12. The method of claim 11, wherein the recipient is one recipient among a plurality of possible recipients, wherein the at least one of the plurality of text prediction uses a plurality of recipient-specific language models, and wherein each recipient-specific language model among the plurality of recipient-specific language models is trained on text previously sent to one recipient among the plurality of possible recipients.

13. The method of claim 11, further comprising:

generating at least one second text prediction, each second text prediction comprising a second term and a second associated probability value, based on the received text input, using the general language model; and

outputting the at least one second text prediction.

14. The method of claim 13, wherein when the term matches the second term, and further comprising retaining the prediction or the second prediction, dependent on whether the associated probability value or the second associated probability value is highest.

15. The method of claim 11, wherein generating the at least one of the plurality of text prediction includes using one or more of an application-specific language model, a purpose or application-type specific language model, a company-specific language model, a user community-specific language model, or a user language model.

16. A text prediction system, comprising:

a processor; and

memory storing instructions that, when executed by the processor, configure the text prediction system to;

receive text input;

identify a location context corresponding to the received text input;

receive a context-specific language model, associated with the location context, trained on prior text inputted from a specific location;

using the context-specific language model associated with the location context, a general language model, and the received text input, concurrently generate a plurality of text predictions, each of the plurality of text predictions comprising a term and an associated probability value; and

output concurrently at least one of the plurality of text predictions including at least one text prediction generated by the context-specific language model for the specific location and at least one text prediction generated by the general language model, wherein the at least one text prediction generated by the general language model is output based on a combined probability corresponding to the at least one text prediction generated by the general language, the combined probability including a context-specific weighting term when the at least one text prediction generated by the general language model is found in the context-specific language model.

17. The text prediction system of claim 16, wherein the instructions further configure the text prediction system to:

generate at least one second text prediction, each second text prediction comprising a second term and a second associated probability value, based on the received text input, using a general language model;

and

output the at least one second text prediction.

18. The text prediction system of claim 17, wherein the at least one text prediction or the at least one second text prediction comprises a word candidate, wherein when a word candidate generated using the general language model matches a word candidate generated using the context-specific language model, the processor is configured to apply a weighting factor to a probability of the word candidate generated using the general language model, and wherein the weighting factor is based on a probability of that word generated using the context-specific language model.

19. The text prediction system of claim 17, wherein the at least one text prediction or the at least one second text prediction comprises a word candidate, and wherein when a word candidate generated using the general language model matches a word candidate generated using the context-specific language model, the text prediction system is configured to calculate a linear combination of the associated probability value generated by the general language model and the associated probability value generated by the context-specific language model for that matching word.

20. The text prediction system of claim 19, wherein the text prediction system calculates a weighted linear combination, wherein the weighting factor is based on a ratio of the associated probability value of that word in the context-specific language model to the second associated probability value of that word in the general language model.