Method and apparatus for automatically invoking a new word module for unrecognized user input

• US 5,852,801 A
• Filed: 10/04/1995
• Issued: 12/22/1998
• Est. Priority Date: 10/04/1995
• Status: Expired due to Term

• Current Assignee: Apple Inc.
• Original Assignee: Apple Computer Incorporated (Apple Inc.)
• Inventors: Hon, Hsiao-Wuen, Chow, Yen-Lu
• Primary Examiner(s): Knepper, David D.
• Assistant Examiner(s): Smits, Talivaldis Ivars
• Application Number:
  US08/538,919
• Time in Patent Office: 1,175 Days
• Field of Search: 395/2.52, 395/2.53, 395/2.79, 704/243, 704/244, 704/270
• US Class Current: 704/244
• CPC Class Codes: G10L 15/005 Language recognition
• CPC Class Codes: G10L 15/18 using natural language modelling
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/197 Probabilistic grammars, e.g. word n-grams
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A method for reducing recognition errors in a speech recognition system that has a user interface, which instructs the user to invoke a new word acquisition module upon a predetermined condition, and that improves the recognition accuracy for poorly recognized words. The user interface of the present invention suggests to a user which unrecognized words may be new words that should be added to the recognition program lexicon. The user interface advises the user to enter words into a new word lexicon that fails to present themselves in an alternative word list for two consecutive tries. A method to improve the recognition accuracy for poorly recognized words via language model adaptation is also provided by the present invention. The present invention increases the unigram probability of an unrecognized word in proportion to the score difference between the unrecognized word and the top one word to guarantee recognition of the same word in a subsequent try. In the event that the score of unrecognized word is unknown (i.e., not in the alternative word list), the present invention increases the unigram probability of the unrecognized word in proportion to the difference between the top one word score and the smallest score in the alternative list.

Claims

1. A computer implemented method for improving the accuracy of a speech recognition system, said method comprising the steps of:

a) detecting a user speech signal;

b) if the user speech signal is not found by the user from an alternative user input candidate list, instructing the user to input the user speech signal again;

c) if the user speech signal is still not found by the user in the alternative user input candidate list, prompting the user to input a representation of the speech signal using a keyboard and invoking a new word module; and

d) adapting a language model using the new word module.

2. The method of claim 1 wherein step (b) further includes the steps of:

i) automatically displaying a top user input candidate in response to the user speech signal; and

ii) displaying the alternative user input candidate list from which a user may select a desired user input; and

wherein step (c) further includes the steps of;

i) automatically displaying a top user input candidate in response to the repeated user speech signal; and

ii) displaying in response to a user command an alternative user input candidate list corresponding to the repeated user speech signal.

3. The method of claim 1 wherein step (a) further includes the steps of:

i) receiving an analog user speech signal;

ii) converting the analog user speech signal into a digital input signal;

iii) and generating a top word candidate and the alternative word list for the digital input signal through a speech recognition program.

4. The method of claim 1 wherein the user input is in the form of sound and the user input program is a speech recognition program.

5. The method of claim 4 wherein the speech recognition program is adapted to recognize a non-alphabetic language.

6. The method of claim 5 wherein the non-alphabetic language is an Asian language.

7. The method of claim 6 wherein the Asian language is Mandarin.

8. The method of claim 1, wherein the speech recognition system comprises a lexicon, said lexicon defining a plurality of words recognized by the system, each word having a score associated with the word for indicating a probability that the word will be the word spoken by a user.

9. The method of claim 8 wherein step (c) further includes the steps of:

a) determining if the unrecognized word is in the lexicon;

b) if the new word is not in the lexicon;

adding this word to the lexicon; and

c) if the unrecognized word is in the lexicon, increasing the score of the unrecognized word by a predetermined amount.

10. The method of claim 9 wherein a predetermined amount is proportional to one of the difference between a top word score and an unrecognized word score and the difference between a top word score and a smallest score in the N-best list.

11. The method of claim 1 wherein step c) further includes

a) instructing the user to invoke a new word module;

b) receiving an open new word module command; and

c) invoking the new word module upon receipt of the command.

12. An apparatus for speech recognition comprising:

an input for receiving speech signals into a processor;

a processor configured to recognize the received speech signals using a language model, wherein if a speech signal is not found by the user from an alternative user input candidate list, the user is instructed to input the speech signal again, wherein if the speech signal is still not found by the user in the alternative user input candidate list, the user is prompted to input a representation of the speech signal using a keyboard and a new word module is invoked, wherein the new word module adapts the language model; and

an output for providing a word sequence representative of the received speech signals.

13. The apparatus of claim 12, wherein the processor is further configured to automatically display a top user input candidate in response to the speech signal and display the alternative user input candidate list from which a user may select a desired user input.

14. The apparatus of claim 12, wherein the processor is further configured to automatically display a top user input candidate in response to the repeated speech signal and display in response to a user command an alternative user input candidate list corresponding to the repeated speech signal.

15. The apparatus of claim 12, wherein the processor is further configured to receive an analog speech signal, convert the analog speech signal into a digital input signal, and generate a top word candidate and the alternative word list for the digital input signal through a speech recognition program.