Deep model statistics method for machine translation
AssigneeABBYY Software Ltd.
InventorsKonstantin Anisimovich, Diar Tuganbaev, Vladimir Selegey, Konstantin Zuev
Life of this patent
Estimating the term…
Abstract
In one embodiment, the invention provides a method for machine translation of a source document in an input language to a target document in an output language, comprising generating translation options corresponding to at least portions of each sentence in the input language; and selecting a translation option for the sentence based on statistics associated with the translation options.
Claims33
A computer method of synthesizing a sentence in a target language from a language independent semantic structure that conveys a meaning of a source sentence in a source natural language, the method comprising:
- gathering a set of candidate linguistic descriptions for linguistic elements of the source sentence, wherein said set of candidate linguistic descriptions include morphological, syntactic, semantic, and pragmatic descriptions related to the linguistic constructions of the source language;
- selecting linguistic descriptions from the set of candidate linguistic descriptions based on ratings, wherein the ratings are based on one or more statistics generated during analyzing a text corpus or a corpus of parallel texts in the source natural language and at least one other language;
- synthesizing a language-independent semantic structure using the selected linguistic descriptions to represent a deep meaning of the source sentence;
- generating, by a processor, a syntactic structure for the sentence in the target language including performing lexical selections on the language-independent semantic structure using linguistic descriptions of the target language based on at least one of the said ratings;
- and
- performing a morphological synthesis on the generated syntactic structure to generate the sentence in the target language using morphological and lexical descriptions of the target language.
A computer system comprising:
- a processor;
- and
- a storage medium coupled to the processor, the storage medium storing instructions which when executed by the processor cause the computer system to:
- gather a set of candidate linguistic descriptions for linguistic elements of the source sentence, wherein said set of candidate linguistic descriptions include morphological, syntactic, semantic, and pragmatic descriptions related to the linguistic constructions of the source language;
- select linguistic descriptions from the set of candidate linguistic descriptions based on ratings, wherein the ratings are based on one or more statistics generated from a text corpus or a corpus of parallel texts in the source natural language and at least one other language;
- synthesize a language-independent semantic structure using the selected linguistic descriptions to represent a deep meaning of the source sentence;
- generate a syntactic structure for the sentence in the target language including performing a lexical selection on the language-independent semantic structure based on at least one of the said ratings;
- and
- performing a morphological synthesis on the generated syntactic structure to generate the sentence in the target language using morphological and lexical descriptions of the target language.
A physical, non-transitory computer storage medium having stored thereon a sequence of instructions which when executed by a computer system cause said computer system to perform a method, the instructions comprising:
- gathering a set of candidate linguistic descriptions for linguistic elements of the source sentence, wherein said set of candidate linguistic descriptions include morphological, syntactic, semantic, and pragmatic descriptions related to the linguistic constructions of the source language;
- selecting linguistic descriptions from the set of candidate linguistic descriptions based on ratings, wherein the ratings are based on one or more statistics generated from a corpus of parallel texts in the source natural language and at least one other language;
- synthesizing a language-independent semantic structure using the selected linguistic descriptions to represent a deep meaning of the source sentence;
- generating a syntactic structure for the sentence in the target language including performing lexical selections on the language-independent semantic structure using linguistic descriptions of the target language based on at least one of the said ratings;
- and
- performing a morphological synthesis on the generated syntactic structure to generate the sentence in the target language using morphological and lexical descriptions of the target language.
Dependencies are inferred from the claim text.