Hybrid baseform generation

• US 7,206,738 B2
• Filed: 08/14/2002
• Issued: 04/17/2007
• Est. Priority Date: 08/14/2002
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Rajput, Nitendra, Verma, Ashish
• Primary Examiner(s): Dorvil; Richemond
• Assistant Examiner(s): SAINT CYR, LEONARD
• Application Number:
  US10/219,686
• Publication Number:
  US 20040034524A1
• Time in Patent Office: 1,707 Days
• Field of Search: 704/9, 704/10, 704/254, 704/260
• US Class Current: 704/10
• CPC Class Codes: G06F 40/242 Dictionaries

Abstract

A method, a computer system and a computer program product for generating baseforms or phonetic spellings from input text are disclosed. The baseforms are initially generated using rules defined for a particular language. Then, phones are identified in the language that are exceptions to the defined rules and an action is associated with each identified phone. A statistical technique is applied to determine whether the identified phones can be modified. Finally, baseforms containing the identified phones that can be modified, are corrected according to the associated actions. Preferably, the statistical technique is only applied to baseforms containing phones that are exceptions to the defined rules. The defined rules can comprise spelling-to-sound rules for a particular phonetic language that incorporate all possible alternative pronunciations of each baseform.

Claims

1. A method for generating baseforms or phonetic spellings front input text, said method comprising:

generating baseforms from said input text using said defined rules for a language;

identifying specific phones, in said language, that are exceptions to said defined rules;

associating actions with said specific phones, wherein each specific phone is associated with only one action that is defined to correct any of said baseforms that contain said specific phone;

identifying each of said baseforms that contain any of said specific phones;

for each of said baseforms that contain said specific phones, applying a statistical technique to determine whether said specific phones can be modified, wherein said statistical technique is applied solely to baseforms containing said specific phones that are exceptions to said defined rules; and

automatically correcting said baseforms containing said specific phones that can be modified by performing said actions that were associated with said specific phones.

2. A method according to claim 1, wherein said defined rules comprise spelling-to-sound rules for a particular phonetic language.

3. A method according to claim 1, wherein all possible alternative pronunciations of each baseform are incorporated in said defined rules.

4. A method according to claim 1, wherein said language comprises Hindi.

5. A method according to claim 1 further comprising applying said statistical technique to reduce a number of alternative baseforms to generate a reduced size pronunciation dictionary.

6. A method according to claim 5, wherein said reduced size pronunciation dictionary is generated by performing only one of one of the following for each one of said specific phones:

removing a phone from a baseform;

replacing a phone by another phone; and

deleting a baseform.

7. A method according to claim 1, wherein said statistical technique uses information relating to the context of a phone in a baseform.

8. A method according to claim 1, wherein for each one of said specific phones said action comprises only one of:

removing a phone from a baseform;

replacing a phone by another phone; and

deleting a baseform.

9. A computer system for generating baseforms or phonetic spellings from input text, said computer system comprising:

a processor means for defining rules for generating a pronunciation dictionary for a particular language;

a processor means for generating baseforms from said input text using said defined rules;

a processor means for identifying specific phones, in said language, that are exceptions to said defined rules;

a processor means for associating actions with said specific phones, wherein each specific phone is associated with only one action that is defined to correct any baseforms that contain said specific phone;

a processor means for identifying each of said baseforms that contain any of said specific phones;

a processor means for applying a statistical technique to each of said baseforms that contain said specific phones in order to determine whether said specific phones can be modified, wherein said statistical technique is applied solely to baseforms containing said specific phones that are exceptions to said defined rules; and

a processor means for automatically correcting said baseforms containing said specific phones that can be modified by performing said actions that were associated with said specific phone.

10. A computer system according to claim 9, wherein said defined rules comprise spelling-to-sound rules for a particular phonetic language.

11. A computer system according to claim 9, wherein all possible alternative pronunciations of each baseform are incorporated in said defined rules.

12. A computer system according to claim 9, wherein said language comprises Hindi.

13. A computer system according to claim 9, further comprising a processor means for applying said statistical technique to reduce a number of alternative baseforms to generate a reduced size pronunciation dictionary.

14. A computer system according to claim 13, further comprising a processor means for generating said reduced size pronunciation dictionary, said processor means for generating said reduced size pronunciation dictionary comprising at least one of:

a processor means for removing a phone from a baseform;

a processor means for replacing a phone by another phone; and

a processor means for deleting a baseform.

15. A computer system according to claim 9, wherein said statistical technique uses information relating to the context of a phone in a baseform.

16. A computer system according to claim 9, wherein for each one of said specific phones said action comprises only one of:

removing a phone from a baseform;

replacing a phone by another phone; and

deleting a baseform.

17. A computer program product having a computer readable medium having a computer program recorded therein for generating baseforms or phonetic spellings from input text, said computer program product comprising:

a computer program code means for defining rules for generating a pronunciation dictionary for a particular language;

a computer program code means for generating baseforms from said input text using said defined rules;

a computer program code means far identifying specific phones, in said language, that are exceptions to said defined rules;

a computer program code means for associating actions with said specific phones, wherein each specific phone is associated with only one action that is defined to correct any baseforms that contain said specific phone;

a computer program code means for identifying each of said baseforms that contain any of said specific phones;

a computer program code means for applying a statistical technique to each of said baseforms that contain said specific phones in order to determine whether said specific phones can be modified, wherein said statistical technique is applied solely to base forms containing said specific phones that are exceptions to said defined rules; and

a computer program code means for automatically correcting said baseforms containing said specific phones that can be modified by performing said actions that were associated with said specific phones.

18. A computer program product according to claim 17, wherein said defined rules comprise spelling-to-sound rules for a particular phonetic language.

19. A computer program product according to claim 17, wherein all possible alternative pronunciations of each baseform are incorporated in said defined rules.

20. A computer program product according to claim 17, wherein said language comprises Hindi.

21. A computer program product according to claim 17, further comprising a computer program code means for applying said statistical technique to reduce a number of alternative baseforms to generate a reduced size pronunciation dictionary.

22. A computer program product according to claim 21, further including processor means for generating said reduced size pronunciation dictionary, said computer program code means comprising at least one of:

a computer program code means for removing a phone from a baseform;

a computer program code means for replacing a phone by another phone; and

a computer program code means for deleting a baseform.

23. A computer program product according to claim 17, wherein said statistical technique uses information relating to the context of a phone in a baseform.

24. A computer program product according to claim 17, wherein for each one of said specific phones said action comprises only one of:

removing a phone from a baseform;

replacing a phone by another phone; and

deleting a baseform.

25. A computer program product having a computer readable medium having a computer program recorded therein for generating baseforms or phonetic spellings from input text, said computer program product including:

a pronunciation dictionary for a particular language;

a list of actions associated with specific phones that are exceptions to rules defined for said language, wherein each specific phone is associated with only one action that is defined to correct any baseforms that contain said specific phone;

a computer program code means for generating baseforms from said input text according to said defined rules;

a computer program code means for identifying baseforms that contain any of said specific phones that are exceptions to said defined rules;

a computer program code means for applying a statistical technique to each of said baseforms that contain said specific phones in order to determine whether said specific phones can be modified, wherein said statistical technique is applied solely to baseforms containing said specific phones that are exceptions to said defined rules; and

computer program code means for automatically correcting said baseforms containing said specific phones that should can be modified by performing said actions that are associated with said specific phones.

26. A computer program product according to claim 25, wherein said statistical technique uses information relating to the context of a phone in a baseform.

27. A computer program product according to claim 25, wherein for each of said specific phones said action comprises only one of:

removing a phone from a baseform;

replacing a phone by another phone; and

deleting a baseform.

28. A computer program product according to claim 25, wherein said defined rules comprise spelling-to-sound rules for a particular phonetic language.

29. A computer program product according to claim 25, wherein said language comprises Hindi.