Method and system for analyzing and translating various languages with use of semantic hierarchy

• US 9,471,562 B2
• Filed: 11/03/2011
• Issued: 10/18/2016
• 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): Pullias, Jesse
• Application Number:
  US13/288,953
• Publication Number:
  US 20120109640A1
• Time in Patent Office: 1,811 Days
• Field of Search: 704 2- 8
• US Class Current: 1/1
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
• 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

A method and computer system for analyzing sentences of various languages and constructing a language-independent semantic structure are provided. On the basis of comprehensive knowledge about languages and semantics, exhaustive linguistic descriptions are created, and lexical, morphological, syntactic, and semantic analyses for one or more sentences of a natural or artificial language are performed. A computer system is also provided to implement, analyze and store various linguistic structures and to perform lexical, morphological, syntactic, and semantic analyses. As result, a generalized data structure, such as a semantic structure, is generated and used to describe the meaning of one or more sentences in language-independent form, applicable to automated abstracting, machine translation, control systems, Internet information retrieval, etc.

Claims

1. A method for analyzing natural language texts using a lexical-semantic hierarchy, the method comprising:

creating, by a computing device, the lexical-semantic hierarchy comprising at least one language-independent semantic class, wherein the semantic class includes one or more semantic attributes and one or more models;

creating at least one language-specific lexical class associated with the lexical-semantic hierarchy, wherein the language-specific lexical class includes one or more semantic and syntactical attributes and one or more models, and wherein the semantic and syntactical attributes and models are inheritable from respective parent units to child units;

creating an instance of the lexical-semantic hierarchy in a memory of the computing device;

identifying a new lexical meaning in a natural language;

identifying at least one candidate position of the new lexical meaning in the lexical-semantic hierarchy in the memory of the computing device on the basis of a dictionary entry for the new lexical meaning;

comparing one or more example sentences comprising the new lexical meaning against one or more semantic and syntactical models of the identified candidate position in the lexical-semantic hierarchy;

creating a new unit in the identified candidate position of the lexical-semantic hierarchy;

creating a syntactical and semantic model for the new unit;

storing the new unit and the created syntactical and semantic model in the memory of the computing device;

receiving a sentence in the natural language; and

representing, using the syntactical and semantic model of the new unit, the sentence in a language-independent form.

2. The method of claim 1, wherein said creating the new unit in the identified candidate position of the lexical-semantic hierarchy is based at least in part upon the comparing of the example sentences comprising the new lexical meaning.

3. The method of claim 1, wherein identifying the candidate position of the new lexical meaning in the lexical-semantic hierarchy further comprises at least one of:

selecting a parent semantic class or creating a new semantic class for the new meaning.

4. The method of claim 1, wherein creating the new unit in the identified candidate position of the lexical-semantic hierarchy further comprises at least one of:

narrowing or splitting the inherited model responsive to detecting a partial mismatch between the example sentence with the new lexical meaning and the one or more semantic and syntactical models, wherein the partial mismatch occurs in one or more slots inherited through the lexical-semantic hierarchy.

5. The method of claim 1, wherein creating the new unit in the identified candidate position of the lexical-semantic hierarchy further comprises restricting one or more models of other units in the lexical-semantic hierarchy where said new unit fills one or more slots of one of more models of the other units.

6. The method of claim 1, further comprising:

creating a plurality of language-specific surface models for syntactic descriptions; and

creating diatheses for the syntactic descriptions to serve as an interface between the language-specific surface models and language-independent deep models of semantic descriptions.

7. The method of claim 6, further comprising:

for each lexical meaning associated with a lexical description, creating a link between the lexical meaning and its language-independent semantic parent, wherein the link indicates a location of a given lexical meaning in the lexical-semantic hierarchy.

8. The method of claim 1, wherein representing the source sentence in the language-independent form comprises determining a language-independent semantic structure of the source sentence.

9. The method of claim 8, further comprising translating the source sentence to a second natural language using the language-independent semantic structure of the source sentence.

10. A computer system, comprising:

a memory for storing data for access by an application program being executed on a computer system, wherein the data relates to a syntactical and semantic model of a natural language, the data comprising a semantic description of entities expressed in the natural language, parametric characteristics of the natural language, and semantic relations with entities of the natural language; and

a processor, coupled to the memory;

wherein the memory is configured to store a data structure, the data structure comprising;

a plurality of language-independent entities represented by a plurality of semantic classes, wherein the semantic classes are arranged into a hierarchy, wherein each of the plurality of semantic classes includes at least one semantic attribute and a semantic model, and wherein the semantic attributes and the semantic models are inherited from parent units to child units;

a plurality of descriptions of semantic relations with the entities, wherein the descriptions of semantic relations are represented by a hierarchical sequence of elements;

a plurality of syntactic descriptions of the natural language, wherein one or more of the plurality of syntactic descriptions are associated with one or more semantic descriptions and one or more lexical descriptions of the natural language;

a plurality of lexical descriptions of the natural language, wherein one or more of the plurality of lexical descriptions of the natural language are associated with one or more of said syntactical descriptions and one or more of said semantic descriptions; and

a plurality of morphological descriptions of the natural language, wherein one or more of the plurality of morphological descriptions of the natural language are associated with one or more of said plurality of lexical descriptions of the natural language, wherein the processor is configured to;

receive a source sentence in a source language,

analyze the source sentence using at least one of the plurality of language-independent entities, and

represent the source sentence in a language-independent form, using at least one of the plurality of assigned descriptions of semantic relations, at least one of the plurality of syntactic descriptions of the natural language, at least one of the plurality of lexical descriptions of the natural language, and at least one of the plurality of morphological descriptions of the natural language.

11. The computer system of claim 10, wherein said plurality of syntactic descriptions of the natural language further comprises:

a plurality of syntactical attributes describing syntactical properties and phenomena of the language;

a plurality of surface models of lexical meanings of the language, wherein each of the plurality of surface models is associated with one or more syntactic descriptions of one or more lexical meanings;

a plurality of analysis rules, wherein each of the plurality of analysis rules is associated with one or more syntactic descriptions of one or more lexical meanings;

a plurality of surface slot descriptions, wherein at least one of the plurality of surface slot descriptions is associated with one or more surface models of one or more lexical meanings; and

a plurality of non-tree syntax descriptions, wherein each non-tree syntax description is associated with one or more element of syntactic descriptions.

12. The computer system of claim 11, wherein each surface model of lexical meanings comprises a plurality of syntactic forms, and wherein said syntactic form further comprises:

a plurality of surface slots, wherein each of the plurality of surface slots is associated with one or more syntactical roles of the lexical meaning as part of lexical description;

a linear order description;

a plurality of diatheses, wherein each diathesis connects a surface slot in the syntactic form with one or more deep slots comprised by a semantic description;

a plurality of grammatical values associated with one or more element of the syntactic form;

a plurality of government and agreement descriptions associated with one or more elements of the syntactic form; and

a plurality of communicative descriptions associated with one or more elements of the syntactic form.

13. The computer system of claim 11, wherein said analysis rules comprises at least one semanteme calculation rule and at least one normalization rule, wherein the semanteme calculation rule is utilized in mapping syntactical properties into semantic properties.

14. The computer system of claim 11, wherein said non-tree syntax description of a natural language comprises a description of an ellipsis description, coordination description, referential description or a structural control description.

15. The computer system of claim 11, wherein the plurality of surface models is a plurality of language-specific descriptions, and wherein diatheses serve as an interface between the language-specific surface models of the syntactic descriptions and language-independent deep models of semantic descriptions.

16. The computer system of claim 10, wherein the semantic description of a natural language further comprises:

a plurality of deep slot descriptions, wherein each of the deep slot reflects a semantic relation in a sentence;

a plurality of semantemes; and

a plurality of pragmatic descriptions.

17. The computer system of claim 10, wherein for each lexical meaning associated with a lexical description, a link is established between the lexical meaning and its language-independent semantic parent, wherein the link indicates a location of a given lexical meaning in a semantic hierarchy.

18. A computer-readable non-transitory storage medium comprising executable instructions that, when executed by a processor, cause the processor to:

create a lexical-semantic hierarchy comprising at least one language-independent semantic class, wherein the semantic class includes one or more semantic attributes and one or more models;

create at least one language-specific lexical class associated with the lexical-semantic hierarchy, wherein the language-specific lexical class includes one or more semantic and syntactical attributes and one or more models, and wherein the semantic and syntactical attributes and models are inheritable from respective parent units to child units;

create an instance of the lexical-semantic hierarchy in the memory;

identify a new lexical meaning in a natural language;

identify at least one candidate position of the new lexical meaning in the lexical-semantic hierarchy in the memory on the basis of a dictionary entry for the new lexical meaning;

compare one or more example sentences comprising the new lexical meaning against one or more semantic and syntactical models of the identified candidate position in the lexical-semantic hierarchy;

create a new unit in position of the lexical-semantic hierarchy;

create a syntactical and semantic model for the new unit;

store, in said memory, the new unit and the created syntactical and semantic model;

receive a sentence in the natural language; and

represent, using the syntactical and semantic model of the new unit, sentence in a language-independent form.