Back to results
US 5,867,811 AGranted
Method, an apparatus, a system, a storage device, and a computer readable medium using a bilingual database including aligned corpora
AssigneesCanon Research Centre Europe Ltd., Canon Europa N.V.
InventorTimothy Francis O'Donoghue
Life of this patent
Estimating the term…
Abstract
Aligned corpora are generated or received from an external source. Each corpus comprises a set of portions aligned with corresponding portions of the other corpus, for example, so that aligned portions are nominally translations of one another in two languages. A statistical database is compiled. An evaluation module calculates correlation scores for pairs of words chosen one from each corpus. Given a pair of text portions (one in each language) the evaluation module combines word pair correlation scores to obtain an alignment score for the text portions. These alignment scores can be used to verify a translation and/or to modify the aligned corpora to remove improbable alignments.
Claims41
A method of operating a processing apparatus for processing a bilingual database, the method comprising the steps:
- storing in the apparatus a database comprising a first and second aligned corpora, each corpus being divided into portions such that at least a subset of the portions of the first corpus are identified by a correspondence relation with respective portions of the second corpus, at least some portions being further divided into plural elementary parts hereinafter referred to as words;
- and
- analyzing the aligned corpora to obtain a statistical database;
- evaluating, using the statistical database, a correlation measure for portions in correspondence relationship chosen from the aligned corpora;
- identifying from the correlation measures of portions in correspondence relationship instances of improbable alignment of the first and second corpora;
- and
- modifying the aligned corpora to remove at least the most improbable corpora alignments.
4 more elements
A method of operating a processing apparatus for processing a bilingual database, the method comprising the steps:
- storing in the apparatus a database comprising first and second aligned corpora, each corpus being divided into portions such that at least a subset of the portions of the first corpus are identified by a correspondence relation with respective portions of the second corpus, at least some portions being further divided into one or more elementary parts hereinafter referred to as words ;
- analyzing the aligned corpora to obtain a statistical database;
- using the statistical database to derive a measure of observed correlation for a plurality of pairs of words, each of the pairs of words chosen one from each corpus;
- receiving a further pair of text portions, one in each of two natural languages;
- evaluating, for the further pair of text portions a measure of correlation of the portions, by combining the derived correlation measures of word pairs contained in the pair of text portions;
- and
- outputting an indication of the likelihood that the further pair of text portions are translations of one another based on the evaluated measure of correlation.
4 more elements
A computer usable medium having computer readable program code means for causing a computer to process a bilingual database, the medium comprising:
- first means for causing the computer to store a database comprising a first and second aligned corpora, each corpus being divided into portions such that at least a subset of the portions of the first corpus are identified by a correspondence relation with respective portions of the second corpus, at least some portions being further divided into plural elementary parts;
- second means for causing the computer to analyze the aligned corpora to obtain a statistical database;
- third means for causing the computer to evaluate, using the statistical database, a correlation measure for portions in correspondence relationship chosen from the aligned corpora;
- fourth means for causing the computer to identify from the correlation measures of portions in correspondence relationship instances of improbable alignment of the first and second corpora;
- and
- fifth means for causing the computer to modify the aligned corpora to remove at least the most improbable corpora alignments.
3 more elements
A computer usable medium having computer readable program code means for causing a computer to automate translation of a source text from a source language to a target language, the medium comprising:
- first means for causing the computer to store a bilingual database including aligned corpora in the source and target languages;
- second means for causing the computer to modify the aligned corpora comprising:
- third means for causing the computer to store a database comprising a first and second aligned corpora, each corpus being divided into portions such that at least a subset of the portions of the first corpus are identified by a correspondence relation with respective portions of the second corpus, at least some portions being further divided into plural elementary parts;
- fourth means for causing the computer to analyze the aligned corpora to obtain a statistical database;
- fifth means for causing the computer to evaluate, using the statistical database, a correlation measure for portions in correspondence relationship chosen from the aligned corpora;
- sixth means for causing the computer to identify from the correlation measures of portions in correspondence relationship instances of improbable alignment of the first and second corpora;
- and
- seventh means for causing the computer to modify the aligned corpora to remove at least the most improbable corpora alignments
- eighth means for causing the computer to partition the source text into portions;
- ninth means for causing the computer to search the aligned corpus of the source language for portions to match the source text portions;
- and
- tenth means for causing the computer to
- (i) for source text portions where a match is found, output the corresponding portion of the target language corpus as a translation for inclusion in an output text, and
- (ii) for portions where no match is found, output a query message indicating that a translation has not been made.
11 more elements
A computer usable medium having computer readable program code means for causing a computer to process a bilingual database, the medium comprising:
- first means for causing the computer to store a database comprising first and second aligned corpora, each corpus being divided into portions such that at least a subset of the portions of the first corpus are identified by a correspondence relation with respective portions of the second corpus, at least some portions being further divided into one or more elementary parts;
- second means for causing the computer to analyze the aligned corpora to obtain a statistical database;
- third means for causing the computer to use the statistical database to derive a measure of observed correlation for a plurality of pairs of words, each of the pairs of words chosen one from each corpus;
- fourth means for causing the computer to receive a further pair of text portions, one in each of two natural languages;
- fifth means for causing the computer to evaluate, for the further pair of text portions a measure of correlation of the portions, by combining the derived correlation measures of word pairs contained in the pair of text portions;
- and
- sixth means for causing the computer to output an indication of the likelihood that the further pair of text portions are translations of one another based on the evaluated measure of correlation.
4 more elements
A computer usable medium having computer readable program code means for causing a computer to automate translation of a source text from a source language to a target language, the medium comprising:
- first means for causing the computer to store a bilingual database comprising aligned corpora in the source and target languages;
- second means for causing the computer to partition the source text into portions;
- third means for causing the computer to obtain a candidate translation of each source text portion in the form of a candidate target language portion;
- and
- fourth means for causing the computer to verify the translation by evaluating a correlation measure for the source language portion and candidate target language portion, said fourth means including;
- fifth means for causing the computer to store a database comprising first and second aligned corpora, each corpus being divided into portions such that at least a subset of the portions of the first corpus are identified by a correspondence relation with respective portions of the second corpus, at least some portions being further divided into one or more elementary parts;
- sixth means for causing the computer to analyze the aligned corpora to obtain a statistical database;
- seventh means for causing the computer to use the statistical database to derive a measure of observed correlation for a plurality of pairs of words, each of the pairs of words chosen one from each corpus;
- eighth means for causing the computer to receive a further pair of text portions, one in each of two natural languages;
- ninth means for causing the computer to evaluate, for the further pair of text portions a measure of correlation of the text portions, by combining the derived correlation measures of word pairs contained in the pair of text portions;
- and
- tenth means for causing the computer to output an indication of the likelihood that the further pair of text portions are likely to be translations of one another based on the evaluated measure of correlation;
- and
- eleventh means for causing the computer to output verified candidate portions for inclusion in a target text.
11 more elements
A processing apparatus for processing a bilingual database, the processing apparatus comprising:
- storage means for storing a database comprising a first and second aligned corpora, each corpus being divided into portions such that at least a subset of the portions of the first corpus are identified by a correspondence relation with respective portions of the second corpus, at least some portions being further divided into plural elementary parts hereinafter referred to as words;
- analyzing means for analyzing the aligned corpora to obtain a statistical database;
- evaluating means for evaluating, using the statistical database, a correlation measure for the portions in correspondence relationship chosen from the aligned corpora;
- identifying means for identifying from the correlation measures of portions in correspondence relationship instances of improbable alignment of the first and second corpora;
- and
- modifying means for modifying the aligned corpora to remove at least the most improbable corpora alignments.
3 more elements
A processing apparatus for processing a bilingual database, the processing apparatus comprising:
- storage means for storing a database comprising first and second aligned corpora, each corpus being divided into portions such that at least a subset of the portions of the first corpus are identified by a correspondence relation with respective portions of the second corpus, at least some of the portions being further divided into one or more elementary parts, hereinafter referred to as words;
- analyzing means for analyzing the aligned corpora to obtain a statistical database;
- deriving means for using the statistical database to derive a measure of observed correlation for a plurality of pairs of words, each pair of words chosen one for each corpus;
- receiving means for receiving a further pair of text portions, one in each of two natural languages;
- evaluating means for evaluating, of the further pair of text portions, a measure of correlation of the portions, by combining the derived correlation measures of word pairs contained in the pairs of text portions;
- and
- outputting means for outputting an indication of the likelihood that the further pair of text portions are translations of one another based on the evaluated measure of correlation.
4 more elements
Dependencies are inferred from the claim text.