CORRECTION OF PREVIOUS WORDS AND OTHER USER TEXT INPUT ERRORS

• US 20160275070A1
• Filed: 03/19/2015
• Published: 09/22/2016
• Est. Priority Date: 03/19/2015
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Mannby, Claes-Fredrik, Corston, Simon, Bradford, Ethan R., McCray, Donni, Unruh, Erland, Kay, David J., Trnka, Keith
• Granted Patent: US 9,760,560 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 3/0236 using selection techniques to select from displayed items
• 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 40/216 using statistical methods
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates

Abstract

The disclosed system provides an efficient method of using a later word or words entered after a previous word along with one or more language models that show which words are likely to occur together to identify a better disambiguated choice for the previous word. To identify the better disambiguated choice for the previous word, the system can evaluate the conditional probability for the later word of various candidate previous words, and select the candidate previous word that has the highest conditional probability. If the conditional probability of the selected candidate previous word exceeds that of the previous word that was entered by at least a factor, then the system can include the selected candidate previous word in a selection list for user selection. The disclosed system also provides an efficient method for using one or more language models and a later word to correct errors in segmenting the word.

Claims

1. A method comprising:

receiving, on a computing device, a plurality of words from a user, the plurality of words including a current word and a previous word preceding the current word;

identifying, based on a language model, a new previous word as having a higher conditional probability of preceding the current word than the previous word that was received from the user;

adding to a selection an entry for back correction suggesting the new previous word as a replacement for the previous word received from the user, the entry including the new previous word together with the current word; and

outputting the selection on the computing device to enable a user selection.

2. The method of claim 1, wherein the language model includes a probability of a previous word or words being followed by a current word.

3. The method of claim 1, further comprising:

receiving a user selection of the entry for back correction; and

in response, replacing the previous word that was received from the user with the new previous word.

4. The method of claim 3, further comprising:

updating a back correction acceptance rate based on the received user selection, wherein the back correction acceptance rate is a ratio of a number of times that the user selects a back correction relative to a word that triggers the back correction from a selection list.

5. The method of claim 1, wherein a conditional probability associated with the new previous word exceeds that of the previous word that was received from the user by at least a factor that is based on an offer rate and an acceptance rate of back corrections.

6. The method of claim 1, wherein the selection also includes an entry for the current word without back correction as the current word was received from the user and a candidate word for replacing the current word, the candidate word being likely to follow the previous word received from the user based on the language model.

7. The method of claim 1, further comprising:

determining that the plurality of words received from the user satisfy a criterion for initiating back correction of the previous word.

8. The method of claim 7, wherein the criterion to be satisfied for initiating back correction includes:

the previous word and the current word are not predictions;

the current word is not a punctuation;

the previous word is not a synthetic word;

or

the previous word is not a word built using two words from the language model.

9. The method of claim 3, further comprising updating probabilities in the language model based on the selection of the entry for back correction.

10. A non-transitory computer-readable medium storing instructions which when executed by a processor causes the processor to:

receive a plurality of words from a user, the plurality of words including a current word and a previous word preceding the current word;

identify, based on a language model, a new previous word as having a higher conditional probability of preceding the current word than the previous word that was received from the user;

add to a selection an entry for back correction suggesting the new previous word as a replacement for the previous word received from the user; and

output the selection to enable a user selection.

16. The computing device of claim 10, wherein the suggested correction that includes the suggested word for the previous word and the current word is identified when there is an ambiguous space between the previous word and the current word.

17. The computing device of claim 16, further configured to identify the suggested correction for the ambiguous space by:

recording a tap sequence in which the previous word and the current word, including the ambiguous space were input by the user; and

identifying, based on the probabilities in the language model, knowledge of a layout of a keypad on the computing device and a context of a word preceding the previous word, the suggested word formed from a combination of the previous word and the current word.

11. A computing device comprising:

a memory; and

a processor disposed to have access to the memory and configured to execute instructions stored in the memory to;

receive multiple words input by a user, the multiple words including a current word and a previous word preceding the current word;

identify, based on probabilities in a language model for predicting a next word that is likely to follow one or more prior words, a suggested correction that includes at least one suggested word to replace the previous word, the current word or both; and

display the suggested correction for user selection.

12. The computing device of claim 11, wherein the at least one suggested word to replace the previous word has a higher likelihood of preceding the current word than the previous word based on the probabilities in the language model.

13. The computing device of claim 11, wherein the suggested correction that includes the suggested words for the previous word and the current word is identified when there is a space transposition error.

14. The computing device of claim 13, further configured to identify the suggested correction for the space transposition error by:

segmenting a tap sequence including the previous word, a space and the current word multiple ways to generate candidates for the previous word and the current word; and

scoring the candidates using the probabilities in the language model to identify candidates for the previous word and the current word that are likely to be words the user intended to input.

15. The computing device of claim 13, further configured to identify the suggested correction for the space transposition error by:

recording a tap sequence in which the previous word and the current word were input by the user; and

searching a language database based on the language model using the tap sequence and the previous word and the current word to identify the suggested words.

18. The computing device of claim 11, wherein the suggested correction that includes the suggested word for the current word is identified when there is a function key tap or a punctuation key tap.

19. A system comprising:

an input interface for accepting user input including a first word followed by a second word;

a back correction suggestion engine configured to;

receive a first selection list including first word candidates for replacing the first word;

determine, based on a language model, a first word candidate that is most likely to occur together with the second word;

an output interface for outputting a selection list that includes the first word candidate for user selection.

20. The system of claim 19, further comprising an integrator for replacing the first word with the first word candidate in response to receiving a user selection of the first word candidate from the selection list via the input interface.