Method, system and program for associating a resource to be translated with a domain dictionary

• US 20030061570A1
• Filed: 09/25/2002
• Published: 03/27/2003
• Est. Priority Date: 09/25/2001
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Hatori, Hiromi, Kamiyama, Yoshiroh
• Granted Patent: US 7,089,493 B2
• Time in Patent Office: Days
• US Class Current: 715/530
• CPC Class Codes: G06F 40/242 Dictionaries
• CPC Class Codes: G06F 40/42 Data-driven translation

Abstract

In a machine translation environment, a context-specific domain directory must sometimes be selected to provide a proper translation in context. A database for trigger patterns associated with specific domains is prepared. When a resource is to be translated, appearances of the pattern within the resource are detected. If appearance frequency for at least one domain exceeds a threshold value, the appearance frequency for that domain is compared to the appearance frequency for other domains. If the appearance frequency for the one domain is found to be significantly higher, the dictionary for that domain is selected for the translation.

Claims

1. For use in a computer system including a database for patterns, each of said patterns being associated with a specific domain in a set of domains, a method for associating a resource to be translated with a domain dictionary comprising the steps of:

calculating, for each domain, an appearance frequency for a pattern included in said resource;

determining, for each domain, whether the appearance frequency of the pattern exceeds a predetermined first threshold value;

where the appearance frequency of the pattern for a particular domain exceeds the predetermined first threshold value, comparing the appearance frequency to the to appearance frequency for the same pattern in other domains to determine whether the frequency for the particular domain is significantly higher; and

where the frequency for the particular domain is significantly higher, selecting the domain dictionary associated with the particular domain for translation of the resource.

2. The method according to claim 1, wherein said resource is a resource on a network having one or more specific address locations on said network;

and wherein said association step is performed by associating an address location with a domain dictionary.

3. The method according to claim 1, wherein said one or more patterns are defined as all or a part of the source words in a domain dictionary.

4. The method according to claim 1, wherein, at said calculating step, a ratio of the number of times a pattern included in said resource to the number of sentences included in said resource is calculated.

5. The method according to claim 1, wherein, at said calculation step, said resource is translated and a ratio of the number of times a pattern is included in the translated portion of said resource to the number of sentences included in said translated portion of said resource is calculated.

6. The method according to claim 1, wherein said calculation step includes a data accumulation step of accumulating data used for calculating the appearance frequency of a pattern;

and wherein the calculation of said appearance frequency of said pattern is performed by using said accumulated data.

7. The method according to claim 6, wherein said data accumulation step includes a step of deleting accumulated data that does not attain a predetermined standard.

8. The method according to claim 7, wherein, at said data accumulation step, said data used for the calculation of an appearance frequency for a pattern is stored in a server that accumulates data generated by multiple users and used to calculate said appearance frequency for said pattern.

9. The method according to claim 1 wherein the comparing step comprises determining whether the ratio of the maximum appearance frequency of a pattern for a domain to the total of the appearance frequencies for said pattern for all the domains exceeds a predetermined second threshold value.

10. The method according to claim 1 comprising the further steps of:

calculating, for each domain, the appearance frequency of a resource that is associated with a specific domain dictionary that is located at a site including said resource;

determining, for each domain, whether said appearance frequency of said resource exceeds a predetermined third threshold value; and

associating said site with a domain dictionary for said domain when it is determined that the appearance frequency for the domain exceeds the predetermined third threshold value.

11. A computer system for associating a resource to be translated with a domain dictionary comprising:

a database of one or more patterns, each of said patterns being associated with a specific domain;

calculation means for calculating, for each domain, an appearance frequency for a pattern included in said resource;

first determination means for determining, for each domain, whether said appearance frequency of said pattern exceeds a predetermined first threshold value;

second determination means for determining whether the maximum appearance frequency for said patterns for a domain for which said first determination means determines that said predetermined first threshold value is exceeded is sharply higher than the appearance frequencies of said pattern for other, different domains; and

association means for, when said second determination means determines that said maximum frequency appearance of said pattern for said domain is sharply higher than said appearance frequencies of said pattern for said different other domains, associating said resource with a domain dictionary for said domain having said maximum appearance frequency of said pattern.

12. The computer system according to claim 11, wherein said resource is a resource on a network having one or more specific address locations on said network;

and wherein said association means associates an address location with a domain dictionary.

13. The computer system according to claim 11, wherein said one or more patterns is defined as all or a part of the source words in a domain dictionary.

14. The computer system according to claim 11, wherein said calculation means calculates a ratio of the number of times a pattern included in said resource to the number of sentences included in said resource.

15. The computer system according to claim 11, wherein said calculation means translates said resource, and calculates a ratio of the number of times a pattern is included in the translated portion of said resource to the number of sentences included in said translated portion of said resource.

16. The computer system according to claim 11, wherein said calculation means includes data accumulation means for accumulating data used for calculating the appearance frequency of a pattern;

and wherein the calculation of said appearance frequency of said pattern is performed by using said accumulated data.

17. The computer system according to claim 16, wherein said data accumulation means includes means for deleting accumulated data that does not attain a predetermined standard.

18. The computer system according to claim 17, wherein said data accumulation means includes means for storing, in a server that accumulates data generated by multiple users to be used for calculation of an appearance frequency for a pattern, said data used for the calculation of said appearance frequency for said pattern.

19. The computer system according to claim 11, wherein said second determination means determines whether the ratio of the maximum appearance frequency of a pattern for a domain to the total of the appearance frequencies for said pattern for all the domains exceeds a predetermined second threshold value.

20. The computer system according to claim 11, further comprising:

calculation means for calculating, for each domain, the appearance frequency of a resource that is associated with a specific domain dictionary that is located at a site including said resource;

third determination means for determining, for each domain, whether said appearance frequency of said resource exceeds a predetermined third threshold value; and

association means for, when it is determined at said third determination step that there is a domain that exceeds said predetermined third value, associating said site with a domain dictionary for said domain.

21. A computer program, for associating a resource to be translated with a domain dictionary permitting a computer that includes a database for one or more patterns, each of said patterns being associated with a specific domain to:

calculate, for each domain, an appearance frequency for a pattern included in said resource;

determine, for each domain, whether said appearance frequency of said pattern exceeds a predetermined first threshold value;

where the appearance frequency for a domain exceeds the predetermined first threshold value, determine whether the maximum appearance frequency is significantly higher than the appearance frequencies of said pattern for other, different domains; and

where the appearance frequency for a domain is significantly higher than the appearance frequency of other domains, associate the resource with a domain dictionary for the.

22. The computer program according to claim 21, wherein said resource is a resource on a network having one or more specific address locations on said network;

and wherein said association process is performed by associating an address location with a domain dictionary.

23. The computer program according to claim 21, wherein said one or more patterns is defined as all or a part of the source words in a domain dictionary.

24. The computer program according to claim 21, wherein, in said calculation process, a ratio of the number of times a pattern included in said resource to the number of sentences included in said resource is calculated.

25. The computer program according to claim 21, wherein, in said calculation process, said resource is translated and a ratio of the number of times a pattern is included in the translated portion of said resource to the number of sentences included in said translated portion of said is calculated.

26. The computer program according to claim 21, wherein said calculation process includes a data accumulation process for accumulating data used for calculating the appearance frequency of a pattern;

and wherein the calculation of said appearance frequency of said pattern is performed by using said accumulated data.

27. The computer program according to claim 26, wherein said data accumulation process includes a process for deleting accumulated data that does not attain a predetermined standard.

28. The computer program according to claim 27, wherein, in said data accumulation process, said data used for the calculation of an appearance frequency for a pattern is stored in a server that accumulates data generated by multiple users and used to calculate said appearance frequency for said pattern.

29. The computer program according to claim 21, wherein said second determination process is performed by determining whether the ratio of the maximum appearance frequency of a pattern for a domain to the total of the appearance frequencies for said pattern for all the domains exceeds a predetermined second threshold value.

30. The computer program according to claim 21, further comprising:

a calculation process for calculating, for each domain, the appearance frequency of a resource that is associated with a specific domain dictionary that is located at a site including said resource;

a third determination process for determining, for each domain, whether said appearance frequency of said resource exceeds a predetermined third threshold value; and

an association process for, when it is determined at said third determination step that there is a domain that exceeds said predetermined third threshold value, associating said site with a domain dictionary for said domain.