Recognition architecture for generating Asian characters

• US 8,457,946 B2
• Filed: 04/26/2007
• Issued: 06/04/2013
• Est. Priority Date: 04/26/2007
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Kuo, Shiun-Zu, Feige, Kevin E., Gong, Yifan, Miwa, Taro, Chitrapu, Arun
• Primary Examiner(s): SAINT CYR, LEONARD
• Application Number:
  US11/789,878
• Publication Number:
  US 20080270118A1
• Time in Patent Office: 2,231 Days
• Field of Search: 704/2, 704/5, 704 7- 10, 704/231, 704/246, 704/251
• US Class Current: 704/2
• CPC Class Codes: G06F 40/129 Handling non-Latin characters, e.g. kana-to-kanji conversion
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation

Abstract

Architecture for correcting incorrect recognition results in an Asian language speech recognition system. A spelling mode can be launched in response to receiving speech input, the spelling mode for correcting incorrect spelling of the recognition results or generating new words. Correction can be obtained using speech and/or manual selection and entry. The architecture facilitates correction in a single pass, rather than multiples times as in conventional systems. Words corrected using the spelling mode are corrected as a unit and treated as a word. The spelling mode applies to languages of at least the Asian continent, such as Simplified Chinese, Traditional Chinese, and/or other Asian languages such as Japanese.

Claims

1. A computer-implemented translation system, comprising:

a speech recognition component for execution by a processor for outputting an Asian character based on speech input, the speech recognition component operable to switch between multiple writing systems based on a language of the speech input; and

a correction component for;

receiving a user selection of a correction method of one of;

a pronunciation method, a character-in-word method, and a character decomposition method to correct an incorrect recognition result; and

generating a corrected output in a single pass using the selected correction method based on an incorrect output by the speech recognition component, wherein the pronunciation method includes a spelling wizard to present one or more dialog panels to facilitate user interaction vocally and manually for selecting and correcting the incorrect output, the one or more dialog panels presented in response to a voice activated command;

wherein the character-in-word method receives information related to how a word is written by from a user voicing words in which the character is used; and

wherein the character decomposition method receives information related to the constituent part with which the character is composed from a user voicing a word to obtain the character by subcomponents.

2. The system of claim 1, wherein the pronunciation method receives information related to word pronunciation using local phonetic symbols.

3. The system of claim 1, further comprising an inference component that employs a probabilistic and/or statistical-based analysis to prognose or infer an action that is desired to be automatically performed.

4. The system of claim 1, further comprising a sharing component for sharing updated lexicon with other users and lexicon vendors.

5. The system of claim 1, further comprising a weighting component for weighting words based on meaning.

6. The system of claim 1, further comprising a frequency component for storing frequency of usage information, the correction component using the frequency of usage information to provide a ranked list of candidate characters.

7. The system of claim 1, wherein the Asian character is of a Chinese or Japanese form of writing that includes a pictograph having an associated fixed pronunciation.

8. A computer-implemented method of recognizing speech, comprising:

recognizing by a processor Asian voice signals to output a result;

switching between multiple writing systems based on a language of the Asian voice signals;

receiving a user selection of a correction method of one of;

a pronunciation method, a character-in-word method, and a character decomposition method to a user to correct an incorrect recognition result;

entering a spelling/correction mode to correct spelling of the result using the selected correction method, wherein the pronunciation method includes a spelling wizard to present one or more dialog panels to facilitate user interaction vocally and manually for selecting and correcting the result, the one or more dialog panels presented in response to a voice activated command, wherein the character-in-word method receives information related to how a word is written from a user voicing words in which the character is used, and wherein the character decomposition method receives information related to the constituent part with which the character is composed from a user voicing a word to obtain the character by subcomponents; and

correcting the spelling of the result in a single pass based on the selected method.

9. The method of claim 8, further comprising automatically entering the spelling/correction mode in response to receiving the Asian voice signals, wherein the spelling wizard receives a phonetic pronunciation.

10. The method of claim 8, further comprising applying a confusion matrix that stores phoneme distances used to correct the spelling.

11. The method of claim 8, further comprising correcting the spelling using a voiced homophone command.

12. The method of claim 8, further comprising correcting the spelling based on word decomposition into symbol subcomponents.

13. The method of claim 8, further comprising developing a new lexicon while correcting the spelling and transmitting the new lexicon to another user.

14. The method of claim 8, further comprising selecting the character based on voice input of one or more words that contain the character.

15. The method of claim 8, further comprising generating a prioritized list of candidate characters and presenting the candidate characters as an indexed list for selection one of manually or by voice command.

16. A computer-implemented system, comprising:

computer-implemented means for recognizing Asian voice signals to output a result;

computer-implemented means for switching between multiple writing systems based on a language of the Asian voice signals;

computer-implemented means for receiving a user selection of a correction method of one of;

a pronunciation method, a character-in-word method, and a character decomposition method to a user, to correct an incorrect recognition result;

computer-implemented means for entering a spelling/correction mode to correct spelling of the result according to the selection, wherein the pronunciation method includes a spelling wizard to present one or more dialog panels to facilitate user interaction vocally and manually for selecting and correcting the result, the one or more dialog panels presented in response to a voice activated command, wherein the character-in-word method receives information related to how a word is written from a user voicing words in which the character is used, and wherein the character decomposition method receives information related to the constituent part with which the character is composed from a user voicing a word to obtain the character by subcomponents;

computer-implemented means for performing a key word search; and

computer-implemented means for correcting the spelling of the result in a single pass based on the selected method.