Translating sentences between languages using language-independent semantic structures and ratings of syntactic constructions

• US 8,195,447 B2
• Filed: 03/22/2007
• Issued: 06/05/2012
• Est. Priority Date: 10/10/2006
• Status: Active Grant

• Current Assignee: ABBYY Development LLC
• Original Assignee: ABBYY Software
• Inventors: Anismovich, Konstantin, Selegey, Vladimir, Zuev, Konstantin
• Primary Examiner(s): Smits, Talivaldis Ivars
• Assistant Examiner(s): ROBERTS, SHAUN A
• Application Number:
  US11/690,102
• Publication Number:
  US 20080086300A1
• Time in Patent Office: 1,902 Days
• Field of Search: 704 1- 10
• 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 represent the meaning of a source sentence from a source language, the method comprising:

obtaining a language-independent semantic structure to represent the meaning of the source sentence;

synthesizing a syntactic structure of an output sentence from the language independent semantic structure using information which includes semantic descriptions, lexical descriptions of an output language, syntactic descriptions of the output language, and morphological descriptions of the output language, wherein the syntactic structure is built at least in part based on ratings of syntactic constructions for each element of the source sentence; and

constructing the output sentence to represent the meaning of the source sentence in the output language by performing a lexical selection on the language-independent semantic structure of the sentence using lexical descriptions of the output language and semantic descriptions, wherein performing the lexical selection further comprises applying one or more semantic structure correction rules to overcome asymmetries between the language-independent semantic structure and the syntactic structure of the output sentence in the output language, and wherein applying one or more semantic structure correction rules to overcome asymmetries includes the use of one or more semanteme calculating rules and one or more semanteme normalization rules.

2. The method of claim 1, further comprising:

determining a linear order and restoring movements on the syntactic structure of the output sentence.

3. The method of claim 1, further comprising:

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

4. The method of claim 1, wherein applying one or more semantic structure correction rules includes transforming a portion of the language-independent semantic structure.

5. The method of claim 1, wherein the morphological synthesis is performed on lexical meanings of constituent cores of the surface structure of the output sentence in the output language.

6. The method of claim 1, wherein the output language comprises a natural language selected from the group consisting of English, French, German, Italian, Russian, Spanish, Ukrainian, Dutch, Danish, Swedish, Finnish, Portuguese, Slovak, Polish, Czech, Hungarian, Lithuanian, Latvian, Estonian, Greek, Bulgarian, Turkish, Tatar, Hindi, Serbian, Croatian, Romanian, Slovenian, Macedonian, Japanese, Korean, Chinese, Arabic, Hebrew, and Swahili.

7. The method of claim 1, wherein obtaining a language-independent semantic structure to represent the meaning of the source sentence includes (1) performing a rough syntactic analysis on the source sentence to generate a graph of generalized constituents and (2) performing a precise syntactic analysis on the graph of the generalized constituents to generate one or more syntactic trees to represent the source sentence from the graph of the generalized constituents.

8. The method of claim 1, wherein the syntactic structure of the sentence includes deep slots and lexical meanings, wherein the language-independent semantic structure is generated by:

(1) substituting each lexical meaning of the sentence in the source language with a corresponding language-independent semantic class, (2) confirming a linear order of lexical meanings, and (3) after the linear order is confirmed, deleting surface slots when generating the language-independent semantic structure, and wherein only deep slots and deep slot descriptions are used to build the language-independent semantic structure of the sentence.

9. The method of claim 1, wherein the one or more semanteme calculating rules are capable of semantically interpreting grammatical values of constituents against a set of semantemes of various semantic categories, and wherein use of the one or more semanteme calculating rules leads to substitution of grammemes with semantemes.

10. A method for a computer system to generate an output sentence in an output language, the method comprising:

performing a lexical selection on a language-independent semantic structure of a source sentence using lexical descriptions of the output language and semantic descriptions, wherein the lexical descriptions and the semantic descriptions are combined into lexical-semantic descriptions, and wherein performing the lexical selection includes applying one or more semantic structure correction rules to overcome asymmetries between the language-independent semantic structure and the syntactic structure of the output sentence in the output language;

building a syntactic structure of the output sentence from the language independent semantic structure using syntactic descriptions and morphological descriptions of the output language, wherein the syntactic structure is built at least in part based on ratings of syntactic constructions for elements of the source sentence;

performing a morphological synthesis on the syntactic structure of the output sentence using morphological descriptions of the output language; and

constructing the output sentence in the output language.

11. The method of claim 10, further comprising:

determining a linear order and restoring movements on the syntactic structure of the output sentence in the output language.

12. The method of claim 10, wherein the morphological synthesis is performed on lexical meanings of constituent cores of the surface structure of the output sentence in the output language.

13. The method of claim 10, wherein the output language comprises a natural language selected from the group consisting of English, French, German, Italian, Russian, Spanish, Ukrainian, Dutch, Danish, Swedish, Finnish, Portuguese, Slovak, Polish, Czech, Hungarian, Lithuanian, Latvian, Estonian, Greek, Bulgarian, Turkish, Tatar, Hindi, Serbian, Croatian, Romanian, Slovenian, Macedonian, Japanese, Korean, Chinese, Arabic, Hebrew, and Swahili.

14. A method for a computer system to generate an output sentence in an output language from a language-independent semantic structure, the method comprising:

performing a lexical selection on the language-independent semantic structure of a source sentence using lexical descriptions of the output language and semantic descriptions, wherein performing the lexical selection includes applying one or more semantic structure correction rules to overcome asymmetries between the language-independent semantic structure and the syntactic structure of the output sentence in the output language;

building a syntactic structure of the output sentence in the output language from the language-independent semantic structure using syntactic descriptions and lexical descriptions of the output language and using a priori rating scores related to a degree of correspondence of precise constituents of the source sentence to corresponding semantic descriptions;

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

performing a morphological synthesis on the syntactic structure of the output sentence in the output language using morphological descriptions of the output language; and

constructing the output sentence in the output language.

15. The method of claim 14, wherein the output language comprises a natural language selected from the group consisting of English, French, German, Italian, Russian, Spanish, Ukrainian, Dutch, Danish, Swedish, Finnish, Portuguese, Slovak, Polish, Czech, Hungarian, Lithuanian, Latvian, Estonian, Greek, Bulgarian, Turkish, Tatar, Hindi, Serbian, Croatian, Romanian, Slovenian, Macedonian, Japanese, Korean, Chinese, Arabic, Hebrew, and Swahili.

16. One or more computer readable media encoded with instructions for causing a computing system to carry out steps comprising:

performing a lexical selection on a language-independent semantic structure using lexical descriptions of an output language and semantic descriptions, wherein performing the lexical selection includes applying one or more semantic structure correction rules to overcome asymmetries between the language-independent semantic structure and the syntactic structure of an output sentence in the output language;

building a surface structure of the output sentence in the output language from the language-independent semantic structure using syntactic descriptions and lexical descriptions of the output language, wherein building the surface structure includes rating a linear order for each syntactic form (“

syntform”

) and creating a precise constituent for each syntform;

performing a morphological synthesis on the surface structure of the output sentence in the output language using morphological descriptions of the output language; and

constructing the output sentence in the output language.

17. The one or more computer readable media of claim 16, wherein the steps further comprising:

determining a linear order and restoring movements on the syntactic structure of the output sentence in the output language.

18. The one or more computer readable media of claim 16, wherein the output language comprises a natural language selected from the group consisting of English, French, German, Italian, Russian, Spanish, Ukrainian, Dutch, Danish, Swedish, Finnish, Portuguese, Slovak, Polish, Czech, Hungarian, Lithuanian, Latvian, Estonian, Greek, Bulgarian, Turkish, Tatar, Hindi, Serbian, Croatian, Romanian, Slovenian, Macedonian, Japanese, Korean, Chinese, Arabic, Hebrew, and Swahili.

19. One or more computer readable media encoded with instructions for causing a computing system to carry out steps for translating a sentence from a source language into an output language, comprising:

analyzing a language-independent semantic structure for the sentence;

performing a lexical selection on the language-independent semantic structure of the sentence using lexical descriptions of the output language and semantic descriptions, wherein performing the lexical selection includes applying one or more semantic structure correction rules to overcome asymmetries between the language-independent semantic structure and a syntactic structure of an output sentence in the output language, wherein the syntactic structure is built at least in part from ratings of syntactic constructions for elements of the source sentence, and wherein applying one or more semantic structure correction rules to overcome asymmetries includes the use of one or more semanteme calculating rules and one or more semanteme normalization rules;

building a surface structure of the output sentence in the output language from the language-independent semantic structure using syntactic descriptions and lexical descriptions of the output language;

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

performing a morphological synthesis on the surface structure of the output sentence in the output language using morphological descriptions of the output language; and

constructing the sentence in the output language.