Translating sentences between languages

• US 8,892,418 B2
• Filed: 07/20/2012
• Issued: 11/18/2014
• Est. Priority Date: 10/10/2006
• Status: Active Grant

• Current Assignee: ABBYY Development LLC
• Original Assignee: ABBYY InfoPoisk LLC
• Inventors: Anisimovich, Konstantin, Selegey, Vladimir, Zuev, Konstantin
• Primary Examiner(s): ROBERTS, SHAUN A
• Application Number:
  US13/554,695
• Publication Number:
  US 20130191109A1
• Time in Patent Office: 851 Days
• Field of Search: 704 1- 9
• US Class Current: 704/4
• CPC Class Codes: G06F 40/40 Processing or translation of natural language natural language analysis G06F40/20; semantic analysis G06F40/30
• CPC Class Codes: G06F 40/56 Natural language generation

Abstract

A method and computer system for translating sentences between languages from an intermediate language-independent semantic representation is provided. On the basis of comprehensive understanding about languages and semantics, exhaustive linguistic descriptions are used to analyze sentences, to build syntactic structures and language independent semantic structures and representations, and to synthesize one or more sentences in a natural or artificial language. A computer system is also provided to analyze and synthesize various linguistic structures and to perform translation of a wide spectrum of various sentence types. As result, a generalized data structure, such as a semantic structure, is generated from a sentence of an input language and can be transformed into a natural sentence expressing its meaning correctly in an output language. The method and computer system can be applied to in automated abstracting, machine translation, natural language processing, control systems, Internet information retrieval, etc.

Claims

1. A method for a computer system to synthesize a sentence into an output language, the method comprising:

obtaining a language-independent semantic structure of the sentence expressed as a language-independent directed acyclic graph including nodes, wherein the nodes correspond to semantic classes;

selecting at least one lexical meaning from multiple lexical meanings in a corresponding semantic class for each node of the acyclic graph using;

a lexical-semantic description of the output language,

a rating related to lexical meanings in the output language,

a rating related to combinations of the lexical meanings in the output language, and

a most number of semantemes assigned to a constituent core of a corresponding word of the sentence, on the basis of the rating related to lexical meanings in the output language and the rating related to combinations of the lexical meanings in the output language;

building a surface structure of the output sentence in the output language from the language-independent semantic structure using a syntactic description of the output language, wherein the surface structure of the output sentence is built at least in part based on a rating of syntactic constructions of the output language;

restoring a movement and determining a linear order on the syntactic structure of the output sentence in the output language; and

performing a morphological synthesis of the output sentence using morphological descriptions of the output language.

2. The method of claim 1, wherein selecting at least one lexical meaning from the multiple lexical meanings in a corresponding semantic class for each node of the acyclic graph is performed by applying semantic structure correction rules to the acyclic graph to overcome asymmetries between the language-independent semantic structure and the syntactic structure of the sentence in the output language.

3. The method of claim 1, wherein building a surface structure of the output sentence in the output language is performed taking into account (1) a referential and structural control description, (2) alternative realization rules and (3) agreement rules.

4. The method of claim 1, wherein building a surface structure of the output sentence in the output language is performed by performing a top-down traversal of the semantic structure.

5. The method of claim 1, wherein building a surface structure of the output sentence in the output language further comprises:

determining surface slots corresponding to deep slots which are presented as arcs in said acyclic graph; and

determining the linear order of words in the output sentence.

6. The method of claim 5, wherein building a surface structure of the output sentence in the output language further comprises restoring movements.

7. The method of claim 5, wherein building a surface structure of the output sentence in the output language further comprises checking structural and referential control.

8. The method of claim 5, wherein determining the linear order of words in the output sentence of the output language is performed taking into account agreement rules.

9. The method of claim 5, wherein determining the linear order of words in the output sentence of the output language is performed taking into account communicative descriptions.

10. A device for translating a text fragment in an input language into an output language, the device comprising:

a processor; and

a memory in electronic communication with the processor, wherein the memory is configured with instructions to cause the processor to perform actions comprising;

obtaining a language-independent semantic structure of the sentence to be synthesized in language-independent terms wherein the meaning of the sentence is expressed as a language-independent directed acyclic graph including nodes, wherein the nodes correspond to semantic classes;

selecting at least one lexical meaning from multiple lexical meanings in a corresponding semantic class for each node of the acyclic graph using;

a lexical-semantic description of the output language,

a rating related to lexical meanings in the output language,

a rating related to combinations of the lexical meanings in the output language, and

a most number of semantemes assigned to a constituent core of a corresponding word of the sentence, on the basis of the rating related to lexical meanings in the output language and the rating related to combinations of the lexical meanings in the output language;

building a surface structure of the output sentence in the output language from the language-independent semantic structure using a syntactic description of the output language, wherein the surface structure of the output sentence is built at least in part based on a rating of syntactic constructions of the output language;

restoring a movement and determining a linear order on the syntactic structure of the output sentence in the output language; and

performing a morphological synthesis of the output sentence using morphological descriptions of the output language.

11. The device of claim 10, wherein selecting at least one lexical meaning from the multiple lexical meanings in a corresponding semantic class for each node of the acyclic graph is performed by applying semantic structure correction rules to the acyclic graph to overcome asymmetries between the language-independent semantic structure and the syntactic structure of the sentence in the output language.

12. The device of claim 10, wherein building a surface structure of the output sentence in the output language is performed taking into account (1) a referential and structural control description, (2) alternative realization rules and (3) agreement rules.

13. The device of claim 10, wherein building a surface structure of the output sentence in the output language is performed by performing a top-down traversal of the semantic structure.

14. The device of claim 10, wherein building a surface structure of the output sentence in the output language further comprises:

determining surface slots corresponding to deep slots which are presented as arcs in said acyclic graph; and

determining the linear order of words in the output sentence.

15. The device of claim 10, wherein building a surface structure of the output sentence in the output language further comprises restoring movements.

16. The device of claim 15, wherein building a surface structure of the output sentence in the output language further comprises checking structural and referential control.

17. The device of claim 15, wherein determining the linear order of words in the output sentence of the output language is performed taking into account agreement rules.

18. The device of claim 15, wherein determining the linear order of words in the output sentence of the output language is performed taking into account communicative descriptions.

19. One or more non-transitory computer-readable media having stored thereon instructions, which when executed by a processing system, cause the system to perform a method for correcting distortions in a scanned image of a page, the method comprising:

obtaining a language-independent semantic structure of the sentence to be synthesized in language-independent terms wherein the meaning of the sentence is expressed as a language-independent directed acyclic graph including nodes, wherein the nodes correspond to semantic classes;

selecting at least one lexical meaning from multiple lexical meanings in a corresponding semantic class for each node of the acyclic graph using;

a lexical-semantic description of the output language,

a rating related to lexical meanings in the output language,

a rating related to combinations of the lexical meanings in the output language, and

a most number of semantemes assigned to a constituent core of a corresponding word of the sentence, on the basis of the rating related to lexical meanings in the output language and the rating related to combinations of the lexical meanings in the output language;

building a surface structure of the output sentence in the output language from the language-independent semantic structure using a syntactic description of the output language, wherein the surface structure of the output sentence is built at least in part based on a rating of syntactic constructions of the output language;

restoring a movement and determining a linear order on the syntactic structure of the output sentence in the output language; and

performing a morphological synthesis of the output sentence using morphological descriptions of the output language.

20. The one or more non-transitory computer-readable media of claim 19, wherein selecting at least one lexical meaning from the multiple lexical meanings in a corresponding semantic class for each node of the acyclic graph is performed by applying semantic structure correction rules to the acyclic graph to overcome asymmetries between the language-independent semantic structure and the syntactic structure of the sentence in the output language.

21. The one or more non-transitory computer-readable media of claim 19, wherein building a surface structure of the output sentence in the output language is performed taking into account (1) a referential and structural control description, (2) alternative realization rules and (3) agreement rules.

22. The one or more non-transitory computer-readable media of claim 19, wherein building a surface structure of the output sentence in the output language is performed by performing a top-down traversal of the semantic structure.

23. The one or more non-transitory computer-readable media of claim 19, wherein building a surface structure of the output sentence in the output language further comprises:

determining surface slots corresponding to deep slots which are presented as arcs in said acyclic graph; and

determining the linear order of words in the output sentence.

24. The one or more non-transitory computer-readable media of claim 23, wherein building a surface structure of the output sentence in the output language further comprises restoring movements.

25. The one or more non-transitory computer-readable media of claim 23, wherein building a surface structure of the output sentence in the output language further comprises checking structural and referential control.

26. The one or more non-transitory computer-readable media of claim 23, wherein determining the linear order of words in the output sentence of the output language is performed taking into account agreement rules.

27. The one or more non-transitory computer-readable media of claim 23, wherein determining the linear order of words in the output sentence of the output language is performed taking into account communicative descriptions.