Claims
- 1. A method in a computer system for correcting misrecognition by a handwriting recognizer, the method comprising the computer-implemented steps of:
- inputting raw handwritten data;
- recognizing the raw handwritten data as a plurality of symbols;
- sending the raw handwritten data and the recognized symbols to a trainer; and
- under control of the trainer,
- displaying the raw handwritten data and the recognized symbols;
- selecting recognized symbols that are misrecognized;
- re-displaying the raw handwritten data corresponding to the selected symbols;
- inputting correct symbols corresponding to the re-displayed raw handwritten data; and
- training the handwriting recognizer to recognize the raw handwritten data corresponding to the selected symbols as the correct symbols.
- 2. A method in a computer system for collecting data for use in training a handwriting recognizer, the handwriting recognizer having a recognition process by which it recognizes handwritten data as symbols, the handwriting recognizer having a training process by which it receives handwritten data and symbols and adjusts its recognition process based on the received handwritten data and the received symbols, the method comprising the computer-implemented steps of:
- collecting handwritten data on a computer handwriting input device;
- displaying the collected handwritten data on a computer output device;
- displaying recognized symbols that correspond to the collected handwritten data on the computer output device;
- selecting a portion of the displayed handwritten data that corresponds to recognized symbols that are misrecognized; and
- inputting at least one correct symbol that the selected portion of the displayed handwritten data should be recognized as by the handwriting recognizer for use in adjusting the recognition process of the handwriting recognizer based on the selected portion of the displayed handwritten data and the input correct symbol.
- 3. The method of claim 2 including the steps of:
- sending the selected portion of the displayed handwritten data and the input correct symbol to the handwriting recognizer for training; and
- adjusting the recognition process based on the sent handwritten data and the sent correct symbol.
- 4. The method of claim 2 or 3 including the step of highlighting the selected portion of the displayed handwritten data.
- 5. The method of claim 2 or 3 including the step of re-displaying the selected portion of the displayed handwritten data out of context from non-selected portions of the displayed handwritten data.
- 6. The method of claim 2 or 3 wherein the input correct symbol is a gesture.
- 7. The method of claim 2 or 3 including the step of recognizing the collected handwritten data as at least one equivalent symbol, wherein a portion of the collected handwritten data is not recognized and wherein a symbol of nonrecognition is generated as the equivalent symbol.
- 8. The method of claim 2 or 3 wherein the step of selecting a portion of the displayed handwritten data includes the steps of:
- recognizing the collected handwritten data as at least one symbol;
- displaying the recognized symbol on a computer output device;
- creating a mapping between each displayed recognized symbol and corresponding displayed handwritten data;
- selecting the displayed recognized symbol that corresponds to a portion of the displayed handwritten data to be selected; and
- using the corresponding displayed handwritten data that maps from the selected displayed recognized symbol as the selected portion of the displayed handwritten data.
- 9. The method of claim 3 including the step of re-recognizing the collected handwritten data after adjusting the recognition process.
- 10. A computer system for training a handwriting recognizer to correctly recognize handwritten data that was previously misrecognized, the computer system comprising:
- a handwriting recognizer for receiving handwritten data and recognizing the handwritten data as symbols;
- a recognizer hook for storing the handwritten data and the recognized symbols; and
- a trainer comprising:
- means for displaying the stored handwritten data and the recognized equivalent symbols;
- means for selecting a portion of the displayed handwritten data that corresponds to recognized symbols that are misrecognized;
- means for inputting a correct symbol that the selected portion of the displayed handwritten data should be recognized as; and
- means for training the handwriting recognizer to recognize the selected portion of the handwritten data as the input correct symbol.
- 11. The system of claim 10 wherein the means for selecting a portion of the displayed handwritten data selects a recognized symbol and selects the handwritten data corresponding to the selected recognized symbol as the selected portion of the displayed handwritten data.
- 12. The system of claim 10 wherein the means for selecting a portion of the displayed handwritten data includes means for identifying handwritten data that is displayed by the display means and selecting a portion of the handwritten data that includes the identified handwritten data such that the selected portion of the displayed handwritten data corresponds to integral recognized symbols.
- 13. The system of claim 10 including means for highlighting the selected portion of the displayed handwritten data.
- 14. The system of claim 10 including means for re-displaying displaying the selected portion of the displayed handwritten data out of context from non-selected portions of the displayed handwritten data.
- 15. The system of claim 10 wherein the means for inputting a correct symbol inputs a gesture.
- 16. A method in a computer system of inputting data for the training of a handwriting recognizer, the method comprising the computer-implemented steps of:
- receiving handwritten data;
- receiving recognized symbols that correspond to the handwritten data;
- receiving a mapping from the recognized symbols to the handwritten data;
- selecting a recognized symbol that is misrecognized;
- displaying a portion of the handwritten data that corresponds to the selected symbol; and
- inputting a correct symbol that the displayed portion of the handwritten data is to be recognized as by the handwriting recognizer for use in training the handwriting recognizer to recognize the displayed portion of the handwritten data as the input correct symbol.
- 17. The method of claim 16 wherein the input symbol is a gesture.
- 18. The method of claim 16 wherein the displayed portion of the handwritten data is displayed out of context from other handwritten data.
- 19. The method of claim 16 including the step of displaying the received handwritten data.
- 20. The method of claim 16 including the step of displaying the received recognized symbols that correspond to the handwritten data.
- 21. A method in a computer system for correcting misrecognition by a handwriting recognizer, the computer system including a memory and a display device, the memory having a database of prototypes with handwritten data and corresponding symbols, the method comprising the computer-implemented steps of:
- inputting raw handwritten data;
- recognizing the raw handwritten data as a plurality of symbols;
- sending the raw handwritten data and the recognized symbols to a trainer; and
- under control of the trainer,
- displaying the raw handwritten data and the recognized symbols on the display device;
- selecting recognized symbols that are misrecognized;
- re-displaying the raw handwritten data corresponding to the selected symbols on the display device;
- inputting correct symbols corresponding to the re-displayed handwritten data;
- training the handwriting recognizer to recognize the raw handwritten data corresponding to the selected symbols as the correct symbols; and
- updating the prototypes in the memory using the raw handwritten data and the corresponding correct symbols.
CROSS-REFERENCE TO RELATED APPLICATION
This application is a continuation of U.S. patent application Ser. No. 07/716,730, filed Jun. 17, 1991, now abandoned.
US Referenced Citations (6)
Foreign Referenced Citations (1)
Number |
Date |
Country |
2187882 |
Jul 1990 |
JPX |
Non-Patent Literature Citations (1)
Entry |
Sklarew et al "Handwritten Data Entry into a Computer", Linus Technologies, Inc. Jun. 1989. |
Continuations (1)
|
Number |
Date |
Country |
Parent |
716730 |
Jun 1991 |
|