Fragment recall and adaptive automated translation

• US 10,635,863 B2
• Filed: 10/30/2017
• Issued: 04/28/2020
• Est. Priority Date: 10/30/2017
• Status: Active Grant

• Current Assignee: SDL PLC (RWS Holdings Plc)
• Original Assignee: SDL PLC (RWS Holdings Plc)
• Inventors: de Vrieze, Erik, Mills, Keith
• Primary Examiner(s): Saint Cyr, Leonard
• Application Number:
  US15/797,974
• Publication Number:
  US 20190129946A1
• Time in Patent Office: 911 Days
• Field of Search: 704 2- 7
• CPC Class Codes: G06F 40/45 Example-based machine translation; Alignment
• CPC Class Codes: G06F 40/58 Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation

Abstract

Fragment recall and adaptive automated translation are disclosed herein. An example method includes determining that an exact or fuzzy match for a portion of a source input cannot be found in a translation memory, performing fragment recall by matching subsegments in the portion against one or more whole translation units stored in the translation memory, and matching subsegments in the portion against corresponding one or more subsegments inside the one or more matching whole translation units, and returning any of the one or more matching whole translation units and the one or more matching subsegments as a fuzzy match, as well as the translations of those subsegments.

Claims

1. A method for automated translation, comprising:

determining that neither an exact match nor a fuzzy match for a portion of a source input in a source language can be found in a translation memory;

performing fragment recall by;

matching subsegments in the portion against one or more whole translation units stored in the translation memory, the matches coming directly from the translation memory;

matching subsegments in the portion against corresponding one or more subsegments inside the one or more matching whole translation units; and

proposing translations of matching subsegments inside the one or more matching whole translation units by utilization of a fine-grained alignment between subsegments of the translation units stored in the translation memory, the fine-grained alignment comprising;

constructing a translation model representing statistical associations between source and target language words; and

using the translation model to compute the most likely alignments between source and target language words in each translation unit in the translation memory; and

returning any of the one or more matching whole translation units and the one or more matching subsegments in a target language as a fuzzy match.

2. The method according to claim 1, further comprising providing the one or more matching subsegments as an autosuggestion in a translation program interface when a user types the source input into the translation program interface.

3. The method according to claim 1, further comprising displaying the one or more matching whole translation units in both a source language version that corresponds to a language of the source input, as well as a target language version.

4. The method according to claim 1, further comprising tracking a number of subsegments utilized from the one or more matching subsegments and the one or more matching whole translation units.

5. The method according to claim 1, further comprising tracking a number of subsegments utilized from the one or more matching subsegments and the one or more matching subsegments inside the one or more matching whole translation units.

6. The method according to claim 1, wherein the fragment recall further comprises utilization of a fine-grained alignment between subsegments of the translation units stored in the translation memory, in order to propose translations of matching subsegments inside the one or more matching whole translation units.

7. The method according to claim 1, further comprising:

evaluating mismatched subsegments;

locating correct translations for the mismatched subsegments; and

replacing the mismatched subsegments with replacement subsegments that correspond to the correct translations.

8. The method according to claim 7, wherein the correct translations are located from a termbase or machine translations.

9. A system of automated translation, the system comprising:

a translation memory;

a translation interface;

a processor; and

a memory for storing executable instructions, the processor executing the instructions to;

determine that neither an exact match nor a fuzzy match for a portion of a source input in a source language can be found in the translation memory;

perform fragment recall by;

matching subsegments in the portion against one or more whole translation units stored in the translation memory, the matches coming directly from the translation memory;

matching subsegments in the portion against corresponding one or more subsegments inside the one or more matching whole translation units; and

proposing translations of the matching subsegments inside the one or more matching whole translation units by utilization of a fine-grained alignment between the matching subsegments of the translation units stored in the translation memory, the fine-grained alignment comprising;

constructing a translation model representing statistical associations between source and target language words; and

using the translation model to compute the most likely alignments between source and target language words in each translation unit in the translation memory; and

return to the translation interface any of the one or more matching whole translation units and one or more of the matching subsegments in a target language as a fuzzy match.

10. The system according to claim 9, wherein the processor further executes the instructions to provide the matching subsegments as an autosuggestion in the translation interface when a user types the source input into the translation interface.

11. The system according to claim 9, wherein the processor further executes the instructions to display the one or more matching whole translation units in both a source language version that corresponds to a language of the source input, as well as a target language version.

12. The system according to claim 9, wherein the processor further executes the instructions to track a number of subsegments utilized from the one or more matching subsegments and the one or more matching whole translation units.

13. The system according to claim 9, wherein the processor further executes the instructions to track a number of subsegments utilized from the one or more matching subsegments and the one or more matching subsegments inside the one or more matching whole translation units.

14. The system according to claim 9, wherein the fragment recall step further comprises utilization of a fine-grained alignment between subsegments of the translation units stored in the translation memory, in order to propose translations of matching subsegments inside the one or more matching whole translation units.

15. The system according to claim 9, wherein the processor further executes the instructions comprising:

evaluating mismatched subsegments;

locating correct translations for the mismatched subsegments; and

replacing the mismatched subsegments with replacement subsegments that correspond to the correct translations.

16. The system according to claim 15, wherein the processor further executes the instructions to obtain correct translations from a termbase or machine translations.