Method and system for automatically determining phonetic transcriptions associated with spelled words

• US 6,233,553 B1
• Filed: 09/04/1998
• Issued: 05/15/2001
• Est. Priority Date: 09/04/1998
• Status: Expired due to Term

• Current Assignee: Panasonic Intellectual Property Corporation of America (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Junqua, Jean-Claude, Contolini, Matteo, Kuhn, Roland
• Primary Examiner(s): Korzuch, William R.
• Assistant Examiner(s): CHAWAN, VIJAY B
• Application Number:
  US09/148,912
• Time in Patent Office: 984 Days
• Field of Search: 704/245, 704/243, 704/256, 704/255, 704/235, 704/257, 704/240, 704/242, 704/231, 704/251, 704/220
• US Class Current: 704/220
• CPC Class Codes: G10L 15/065 Adaptation
• CPC Class Codes: G10L 2015/086 Recognition of spelled words

Abstract

New entries are added to the lexicon by entering them as spelled words. A transcription generator, such as a decision-tree-based phoneme or morpheme transcription generator, converts each spelled word into a set of n-best transcriptions or sequences. Meanwhile, user input or automatically generated speech corresponding to the spelled word is processed by an automatic speech recognizer and the recognizer rescores the transcriptions or sequences produced by the transcription generator. One or more of the highest scored (highest confidence) transcriptions may be added to the lexicon to update it. If desired, the spelled word-pronunciation pairs generated by the system can be used to retrain the transcription generator, making the system adaptive or self-learning.

Claims

1. A method for automatically generating the phonetic transcription associated with a spelled word, comprising:

transcribing said spelled word into sound units to generate a plurality of transcriptions each corresponding to said spelled word without using a pre-existing dictionary;

associating a score with each transcription;

supplying said plurality of transcriptions to an automatic speech recognizer;

supplying speech data corresponding to said spelled word to said automatic speech recognizer when none of the scores is above a predetermined threshold; and

using said automatic speech recognizer to rescore said transcriptions based on said speech data.

2. The method of claim 1 wherein said transcribing step generates a plurality of phonetic transcriptions.

3. The method of claim 1 wherein said transcribing step generates a plurality of morphemic transcriptions.

4. The method of claim 1 wherein said sound units correspond to acoustic models.

5. The method of claim 1 wherein said sound units correspond to speech templates.

6. The method of claim 1 further comprising using said recognizer to select at least one transcription for updating a lexicon.

7. The method of claim 1 wherein said transcribing step is performed using a trainable transcription generator and wherein said method further comprises using said recognizer to select at least one transcription and using said at least one transcription to retrain said transcription generator.

8. The method of claim 1 wherein said transcribing step is performed using a trainable transcription generator employing at least one decision tree and wherein said method further comprises using said recognizer to select at least one transcription and using said at least one transcription to update said at least one decision tree.

9. The method of claim 1 further comprising selecting at least one of said transcriptions and using said at least one transcription to update a lexicon.

10. A system for updating a lexicon based on spelled word input comprising:

transcription generator receptive of said spelled word input for generating a plurality of scored transcriptions without using a pre-existing dictionary; and

a confidence checking mechanism for determining whether any of the scored transcriptions has a confidence level above a predetermined threshold;

an automatic speech recognizer receptive of speech data corresponding to said spelled word input for rescoring said plurality of scored transcriptions to generate a plurality of rescored transcriptions when none of the scored transcriptions has a confidence level above the predetermined threshold; and

selection mechanism for selecting and using at least one of said rescored transcriptions to update said lexicon.

11. The system of claim 10 wherein said transcription generator produces a set of phonetic transcriptions.

12. The system of claim 10 wherein said transcription generator produces a set of morpheme transcriptions.

13. The system of claim 10 wherein said transcription generator is a phoneticizer employing decision trees.

14. The system of claim 10 wherein said selection mechanism provides at least one of said rescored transcriptions for retraining said transcription generator.

15. The system of claim 10 wherein said transcription generator is a phoneticizer employing decision trees and wherein said selection mechanism provides at least one of said rescored transcriptions for updating said decision trees.