Confusion matrix mediated word prediction

• US 5,963,666 A
• Filed: 08/18/1995
• Issued: 10/05/1999
• Est. Priority Date: 08/18/1995
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Srinivasan, Savitha, Fujisaki, Tetsunosuke
• Primary Examiner(s): CHANG, JON CARLTON
• Application Number:
  US08/516,785
• Time in Patent Office: 1,509 Days
• Field of Search: 382/187, 382/309, 382/310, 382/225, 382/228, 382/229, 382/177, 382/186, 382/230, 382/217, 382/218, 707/534
• US Class Current: 382/187
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/268 Lexical context

Abstract

A system and method for performing an incremental search using a character string returned by a recognizer on a confusion matrix encoded dictionary to predict a word in a handwriting input field of an application program. In the system and method, handwritten characters are input into a data entry field of an application program. The handwritten characters are recognized and assigned a cluster code. The string of characters is characterized by the cluster codes of the individual characters. The cluster codes of the string are compared with strings of cluster codes representing words in a dictionary. All or some of the matching words are displayed to the user, from which a selection can be made.

Claims

1. A system for providing user generated data to an application computer program, comprising:

means for translating a string of handwritten characters representing the user data into digitized handwriting signals, said string of handwritten characters having been previously generated by a character recognition system;

means for comparing the digitized handwriting signals with electronically stored prototype handwriting signals representing prototype handwritten characters and for assigning a closest matching character identifier to each of the digitized handwriting signals;

means for assigning to each closest matching character identifier one or more cluster codes, the cluster codes representing clusters of confusable characters containing the closest matching character;

means for providing, for each combination of cluster codes representing the string of handwritten characters, a string of cluster codes representing the string of handwritten characters;

means for comparing the strings of cluster codes with cluster code strings representing words in a dictionary;

means for generating and displaying a list of words in the dictionary having cluster code strings matching one or more of the strings of cluster codes representing the string of handwritten characters;

means for selecting one of the words from the list; and

means for providing the selected word to the application program.

2. The system of claim 1, wherein two or more of the clusters contain a common character.

3. The system of claim 1, wherein the clusters contain no overlapping characters.

4. The system of claim 1, further comprising means for pruning the generated list to remove unlikely matches prior to presenting it to a user.

5. The system of claim 1, further comprising means for ranking the list in order of likelihood of matching the string of closest matching character identifiers associated with the handwritten characters.

6. The system of claim 5, wherein the means for ranking comprises:

means for comparing each character in the string of closest matching characters with its corresponding character in each dictionary word in the generated list and assigning a score based upon the number of character matches or cluster matches and the cohesiveness of the cluster;

means for ordering the generated list based upon the score; and

means for removing from the ordered list all but the first n words.

7. The system of claim 5, wherein the string of handwritten characters is less than all of the characters in a word being entered, and wherein the means for generating and displaying generates and displays complete words beginning with the string of handwritten characters.

8. The system of claim 1, further comprising:

means for creating the clusters of confusable characters; and

means for encoding a raw dictionary with the cluster codes that represent the character confusions to produce the cluster code strings representing words in a dictionary;

wherein a different clustering scheme may be selected for the encoding of the dictionary for each character.

9. The system of claim 1, wherein the cluster codes represent the cohesiveness of the clusters of confusable characters.

10. The system of claim 1, wherein the clusters are broad, such that each cluster contains many characters, resulting in few clusters.

11. The system of claim 1, wherein the clusters are narrow, such that each cluster contains few characters, resulting in many clusters.

12. A method for providing user generated data to an application computer program, comprising:

translating a string of handwritten characters representing the user data into digitized handwriting signals, said string of handwritten characters having been previously generated by a character recognition system;

comparing the digitized handwriting signals with electronically stored prototype handwriting signals representing prototype handwritten characters and assigning a closest matching character identifier to each of the digitized handwriting signals;

assigning to each closest matching character identifier one or more cluster codes, the cluster codes representing clusters of confusable characters containing the closest matching character;

providing, for each combination of cluster codes representing the string of handwritten characters, a string of cluster codes representing the string of handwritten characters;

comparing the strings of cluster codes with cluster code strings representing words in a dictionary;

generating and displaying a list of words in the dictionary having cluster code strings matching one or more of the strings of cluster codes representing the string of handwritten characters;

selecting one of the words from the list; and

providing the selected word to the application program.

13. The method of claim 12, wherein two or more of the clusters contain a common character.

14. The method of claim 12, wherein the clusters contain no overlapping characters.

15. The method of claim 12, further comprising pruning the generated list to remove unlikely matches prior to presenting it to a user.

16. The method of claim 12, further comprising the step of ranking the list in order of likelihood of matching the string os closest matching character identifiers associated with the handwritten characters, said step of ranking further comprising the steps of:

comparing each character in the string of closest matching characters with its corresponding character in each dictionary word in the generated list and assigning a score based upon the number of matches;

ordering the generated list based upon the score; and

removing from the ordered list all but the first n words.

17. The method of claim 16, wherein the string of handwritten characters is less than all of the characters in a word being entered, and wherein the step of generating and displaying generates and displays complete words beginning with the string of handwritten characters.

18. A system for providing user generated data to an application computer program, comprising:

means for translating a string of handwritten characters representing the user data into digitized handwriting signals, said string of handwritten characters having been previously generated by a character recognition system;

means for comparing the digitized handwriting signals with electronically stored prototype handwriting signals representing prototype handwritten characters and assigning a closest matching character identifier to each of the digitized handwriting signals;

means for assigning to each closest matching character identifier one or more cluster codes, the cluster codes representing clusters of confusable characters containing the closest matching character;

means for providing, for each combination of cluster codes representing the string of handwriting characters, a string of cluster codes representing the string of handwritten characters;

means for comparing the strings of cluster codes with cluster code strings representing words in a dictionary;

means for generating and displaying a list of words in the dictionary having cluster code strings matching one or more of the strings of cluster codes representing the string of handwritten characters;

means for pruning the generated list to remove unlikely matches prior to presenting it to the user;

means for selecting one of the words from the list; and

means for providing the selected word to the application program.

19. A stored program device readable by a computer, having program instructions executable by the computer to perform method steps for handwriting recognition, the method steps comprising:

translating a string of handwritten characters representing the user data into digitized handwriting signals, said string of handwritten characters being generated by a character recognition system prior to said step of translating;

assigning an identifier to each of the digitized handwriting signals;

assigning at least one cluster code to each identifier, said cluster code represents clusters of confusable characters containing a closest matching character associated with said identifier;

providing a string of cluster codes representing the string of handwritten characters;

comparing the string of cluster codes with cluster code strings representing words in a dictionary;

generating a list of words in the dictionary having cluster code strings matching at least one of the string of cluster codes representing the string of handwritten characters;

ranking the generated list in order of likelihood of matching a string of closest matching character identifiers associated with the handwritten characters;

selecting one of the words from the list; and

providing the selected word to the application program.

20. The method according to claim 19, further comprising the step of comparing the digitized handwriting signals with electronically stored prototype handwriting signals representing prototype handwritten characters.

21. The method according to claim 19, further comprising the step of pruning the list to remove unlikely matches prior to displaying it to the user.