Method for translating documents from one language into another using a database of translations, a terminology dictionary, a translation dictionary, and a machine translation system

• US 8,548,795 B2
• Filed: 08/06/2008
• Issued: 10/01/2013
• Est. Priority Date: 10/10/2006
• Status: Expired due to Fees

• Current Assignee: ABBYY Development LLC
• Original Assignee: ABBYY Software
• Inventors: Anisimovich, Konstantin, Selegey, Vladimir, Zuev, Konstantin
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): ADESANYA, OLUJIMI A
• Application Number:
  US12/187,131
• Publication Number:
  US 20090070099A1
• Time in Patent Office: 1,882 Days
• Field of Search: 704/2, 704/5, 704/7, 704/8, 704/9
• US Class Current: 704/4
• CPC Class Codes: G06F 40/268 Morphological analysis
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06F 40/55 Rule-based translation

Abstract

In one embodiment, the invention provides a method for translating a document in an input language into an output language comprising: a) for each document fragment for which a translation is readily available, translating said document fragment based on said readily available translation; and b) for each remaining untranslated fragment for which a translation is not readily available, translating said untranslated fragment based on a model-based machine translation technique. A translation is readily available if a search reveals at least one matching translation for the document fragment in a translation database.

Claims

1. A method for translating a document in an input language into an output language, the method comprising:

a) segmenting the document into fragments;

b) for each document fragment for which a translation is readily available, translating said document fragment based on said readily available translation; and

c) for each remaining untranslated document fragment for which a translation is not readily available, translating said untranslated document fragment based on a model-based machine translation technique by;

generating, using a computer processor, a graph of generalized constituents from a lexical-morphological structure of the untranslated document fragment by using one or more syntactical descriptions, semantic descriptions or lexical descriptions of the input language;

generating one or more syntactic trees from the graph of generalized constituents;

generating one or more rating scores for the one or more syntactic trees;

using linguistic descriptions of the input language or output language, and the one or more rating scores, building a language-independent semantic structure to represent the meaning of each untranslated document fragment; and

providing syntactically coherent output based on the respective language-independent semantic structure, wherein syntactically coherent output includes substitutions that agree both between themselves and with other words in the syntactically coherent output, morphologically and syntactically.

2. The method of claim 1, wherein steps b) and c) are performed automatically.

3. The method of claim 2, wherein a translation is readily available if a search reveals at least one matching translation for the document fragment in a translation database.

4. The method of claim 3, further comprising performing the search.

5. The method of claim 4, wherein the search reveals at least one matching translation.

6. The method of claim 5, wherein the search is an exact search or wherein each matching translation corresponds exactly to the document fragment.

7. The method of claim 5, wherein the search is a fuzzy search where each matching translation does not correspond exactly to the document fragment and includes at least one differing word that is not in the document fragment.

8. The method of claim 6, wherein said translating comprises substituting one matching translation for the document fragment.

9. The method of claim 7, wherein said translating comprises substituting one matching translation for the document fragment based on the matching translation and substituting each differing word in the matching translation based on a lookup of the differing word in at least one of a terminology dictionary and a translation dictionary.

10. The method of claim 8, or claim 9, wherein the one matching translation is the matching translation in the case of a single matching translation or a selected matching translation from a set of translation variants in the case of more that one matching translation.

11. The method of claim 10, further comprising selecting the matching translation variant from the set of translation variants.

12. The method of claim 11, wherein said selecting is performed by a machine.

13. The method of claim 11, wherein said selecting is performed with assistance from a human operator.

14. The method of claim 12, further comprising providing a user input interface and allowing a human to edit the selected translation variant through the user input interface.

15. The method of claim 1, wherein said segmenting is performed by a machine.

16. The method of claim 15, further comprising displaying the fragments on a user input interface and allowing a human operator to edit the document fragments generated by said segmenting.

17. The method of claim 1, wherein said segmenting is performed with human assistance.

18. The method of claim 3, further comprising performing one of creating and selecting the translation database.

19. The method of claim 9, further comprising performing one of creating and selecting the terminology dictionary for a given subject domain.

20. The method of claim 9, further comprising performing one of creating and selecting the translation dictionary for a given subject domain.

21. The method of claim 1, wherein a fragment is one or more word combination, part of a sentence, sentence or paragraph.

22. The method of claim 1, wherein each semantic structure is built with semantic descriptions, wherein a language-independent semantic description includes at least a language-independent description of a deep constituent or a deep slot.

23. The method of claim 1, wherein the language-independent semantic structure is based at least in part upon a syntactic structure of the document fragment in the input language, and wherein the input language is a natural language.

24. A system comprising:

a processor; and

a memory coupled to the processor, the memory comprising a translation application which when executed causes the system to perform a set of instructions for translating a document in an input language into an output language, the instructions comprising;

a) segmenting the document into document fragments;

b) for each document fragment for which a translation is readily available, translating said document fragment based on said readily available translation; and

c) for each remaining untranslated document fragment for which a translation is not readily available, translating said untranslated document fragment based on a model-based machine translation technique by;

generating a graph of generalized constituents from a lexical-morphological structure of the untranslated document fragment by using one or more syntactical descriptions, semantic descriptions or lexical descriptions of the input language;

generating one or more syntactic trees from the graph of generalized constituents;

generating one or more rating scores for the one or more syntactic trees;

using linguistic descriptions of the input language or output language, and the one or more rating scores, building a language-independent semantic structure to represent the meaning of each untranslated document fragment; and

providing syntactically coherent output based at least in part upon the respective language-independent semantic structure, wherein syntactically coherent output includes substitutions that agree both between themselves and with other words in the syntactically coherent output, morphologically and syntactically.

25. The system of claim 24, wherein steps b) and c) are performed automatically.

26. The system of claim 25, wherein a translation is readily available if a search reveals at least one matching translation for the document fragment in a translation database.

27. The system of claim 26, wherein the instructions further comprise performing the search.

28. The system of claim 27, wherein the search reveals at least one matching translation.

30. The system of claim 28, wherein the search is a fuzzy search where each matching translation does not correspond exactly to the document fragment and includes at least one differing word that is not in the document fragment.

32. The system of claim 30, wherein said translating comprises substituting one matching translation for the document fragment based on the matching translation and substituting each differing word in the matching translation based on a lookup of the differing word in at least one of a terminology dictionary and a translation dictionary.

29. The system of system 28, wherein the search is an exact search or each matching translation corresponds exactly to the document fragment.

31. The system of claim 29, wherein said translating comprises substituting one matching translation for the document fragment.

33. The system of claim 31, or claim 32, wherein the one matching translation is the matching translation in the case of a single matching translation or a selected matching translation from a set of translation variants in the case of more than one matching translation.

34. The system of claim 33, further comprising selecting the matching translation variant from the set of translation variants.

35. The system of claim 34, wherein said selecting is performed by a machine.

36. The system of claim 34, wherein said selecting is performed with assistance from a human operator.

37. A physical, non-transitory computer storage medium having stored thereon a program which when executed by a processor, perform instructions for translating a document in an input language into an output language, the instructions comprising:

a) segmenting the document into document fragments;

b) for each document fragment for which a translation is readily available, translating said document fragment based on said readily available translation; and

c) for each remaining document untranslated fragment for which a translation is not readily available, translating said untranslated document fragment based on a model-based machine translation technique by;

generating a graph of generalized constituents from a lexical-morphological structure of the untranslated document fragment by using one or more syntactical descriptions, semantic descriptions or lexical descriptions of the input language;

generating one or more syntactic trees from the graph of generalized constituents;

generating one or more rating scores for the one or more syntactic trees;

using linguistic descriptions of the input language or output language, and the one or more rating scores, building a language-independent semantic structure to represent the meaning of each untranslated document fragment; and

providing syntactically coherent output based at least in part upon the respective language-independent semantic structure.

38. The computer storage medium of claim 37, wherein steps b) and c) are performed automatically.

39. The computer storage medium of claim 38, wherein a translation is readily available if a search reveals at least one matching translation for the document fragment in a translation database.