Method, non-transitory computer-readable recording medium storing a program, apparatus, and system for creating similar sentence from original sentences to be translated

• US 10,303,761 B2
• Filed: 09/07/2017
• Issued: 05/28/2019
• Est. Priority Date: 09/27/2016
• Status: Active Grant

• Current Assignee: Panasonic Intellectual Property Management Co., Ltd. (Panasonic Holdings Corporation)
• Original Assignee: Panasonic Intellectual Property Management Co., Ltd. (Panasonic Holdings Corporation)
• Inventors: Yamauchi, Masaki, Fujiwara, Nanami, Imade, Masahiro
• Primary Examiner(s): Sharma, Neeraj
• Application Number:
  US15/697,489
• Publication Number:
  US 20180089169A1
• Time in Patent Office: 628 Days
• Field of Search: None
• CPC Class Codes: G06F 40/151 Transformation
• CPC Class Codes: G06F 40/247 Thesauruses; Synonyms
• CPC Class Codes: G06F 40/268 Morphological analysis
• CPC Class Codes: G06F 40/45 Example-based machine translation; Alignment
• CPC Class Codes: G06F 40/49 using very large corpora, e.g. the web

Abstract

In a method of creating similar sentences from an entered original, one or more second phrases having the same meaning as a first phrase, which is part of the original, are extracted from a first database; an N-gram value is calculated according to a context dependence value, in a second database, corresponding to the one or more second phrases; one or more contiguous third phrases that include a number of second phrases equivalent to the N-gram value are extracted from one or more sentences obtained by replacing, in the original, the first phrase with the one or more second phrases; the appearance frequency of the one or more third phrases in a third database is calculated; and if the calculated appearance frequency is determined to be larger than or equal to a threshold, the one or more sentences are used as similar sentences of the original and are externally output.

Claims

1. A method of creating similar sentences from an original sentence to be translated, the method comprising:

accepting, by a processor, a first sentence including a first phrase;

extracting, from a first database by the processor, one or more second phrases having the same meaning as the first phrase, the first phrase being part of a plurality of phrases constituting the first sentence, the first database associating phrases and synonyms of the phrases with each other;

calculating, by the processor, an N-gram value according to a context dependence value corresponding to the one or more second phrases, the context dependence value being obtained from a second database, the second database associating phrases and context dependence values, corresponding to the phrases included in the second database, with each other, the context dependence value indicating a degree to which a meaning of a phrase included in the second database depends on context;

extracting, by the processor, one or more contiguous third phrases that include a number of second phrases equivalent to the N-gram value from one or more second sentences obtained by replacing, in the first sentence, the first phrase with the one or more second phrases;

calculating, by the processor, an appearance frequency of the one or more third phrases in a third database, the third database associating phrases and appearance frequencies of the phrases, in the third database, with each other;

determining, by the processor, whether the calculated appearance frequency is larger than or equal to a threshold;

determining, by the processor, if the calculated appearance frequency is determined to be larger than or equal to the threshold, the one or more second sentences as a substitute of the first sentence;

outputting, by the processor, the one or more second sentences as the substitute, to an external device;

generating, by the processor, an updated translation model based on the determination of the substitute; and

perform, by the processor, a machine translation using the updated translation model.

2. The method according to claim 1, wherein:

the first sentence is written in a first language;

the first sentence is included in a parallel corpus, the parallel corpus including a plurality of pairs, each of which is composed of a sentence written in the first language and a translated sentence written in a second language; and

if the calculated appearance frequency is determined to be larger than or equal to the threshold, the one or more second sentences are added to the parallel corpus as similar sentences of the first sentence.

3. The method according to claim 1, wherein:

the third database includes an N-gram language model database;

i (positive integer) is determined as N of the N-gram language model according to the context dependence value;

the third database is referenced to obtain an appearance frequency of the N-gram including the second phrase; and

whether to use the one or more second sentences as similar sentences of the first sentence is determined according to the appearance frequency of the N-gram including the second phrase.

4. The method according to claim 1, wherein:

a translation result sentence is created by translating a predetermined to-be-translated sentence by use of the updated translation model, wherein the updated translation model is generated from the one or more second sentences determined to be used as similar sentences of the first sentence and from a translated sentence resulting from translation of the first sentence, from which the one or more second sentences have been created, in a second language;

the translation result sentence is evaluated; and

feedback information is created according to an evaluation result for the translation result sentence, the feedback information including language information related to at least one of a language of the to-be-translated sentence and a language of the translation result sentence as well as evaluation information for the language information.

5. The method according to claim 4, wherein at least one of the first database, the second database, and the third database is updated with the feedback information.

6. The method according to claim 4, wherein if the feedback information includes the second phrase with context dependence, the second database and the third database are updated.

7. The method according to claim 4, wherein if the feedback information includes a new sentence representation, a context dependence value in the second database is changed according to the new sentence representation.

8. The method according to claim 4, wherein if the feedback information includes a new sentence representation, the third database is updated so as to include the new sentence representation.

9. A non-transitory computer-readable recording medium storing a program that causes a computer to function as an apparatus that creates similar sentences from an original sentence to be translated, wherein the program causes the computer to execute processing to:

accept a first sentence including a first phrase;

extract, from a first database, one or more second phrases having the same meaning as the first phrase, the first phrase being part of a plurality of phrases constituting the first sentence, the first database associating phrases and synonyms of the phrases with each other;

calculate an N-gram value according to a context dependence value corresponding to the one or more second phrases, the context dependence value being obtained from a second database, the second database associating phrases and context dependence values, corresponding to the phrases included in the second database, with each other, the context dependence value indicating a degree to which a meaning of a phrase included in the second database depends on context;

extract one or more contiguous third phrases that include a number of second phrases equivalent to the N-gram value from one or more second sentences obtained by replacing, in the first sentence, the first phrase with the one or more second phrases;

calculate an appearance frequency of the one or more third phrases in a third database, the third database associating phrases and appearance frequencies of the phrases, in the third database, with each other;

determine whether the calculated appearance frequency is larger than or equal to a threshold; and

determine, if the calculated appearance frequency is determined to be larger than or equal to the threshold, the one or more second sentences as a substitute of the first sentence;

output the one or more second sentences as the substitute, to an external device;

generate an updated translation model based on the determination of the substitute; and

perform a machine translation using the updated translation model.

10. An apparatus that creates similar sentences from an original sentence to be translated, the apparatus comprising:

an acceptor that accepts a first sentence including a first phrase;

a second phrase extractor that extracts, from a first database, one or more second phrases having the same meaning as the first phrase, the first phrase being part of a plurality of phrases constituting the first sentence the first database associating phrases and synonyms of the phrases with each other;

a calculator that calculates an N-gram value according to a context dependence value corresponding to the one or more second phrases, the context dependence value being obtained from a second database, the second database associating phrases and context dependence values, corresponding to the phrases included in the second database, with each other, the context dependence value indicating a degree to which a meaning of a phrase included in the second database depends on context;

a third phrase extractor that extracts one or more contiguous third phrases that include a number of second phrases equivalent to the N-gram value from one or more second sentences obtained by replacing, in the first sentence, the first phrase with the one or more second phrases;

a calculator that calculates an appearance frequency of the one or more third phrases in a third database, the third database associating phrases and appearance frequencies of the phrases, in the third database, with each other;

a determiner that determines whether the calculated appearance frequency is larger than or equal to a threshold;

an outputer that, if the calculated appearance frequency is determined to be larger than or equal to the threshold, uses the one or more second sentences as a substitute of the first sentence, and outputs the one or more second sentences as the substitute, to an external device;

a translation model creator that generates an updated translation model based on the determination of the substitute; and

a machine translator that performs a machine translation using the updated translation model.

11. A system that creates similar sentences from an original sentence to be translated, the system comprising:

the apparatus according to claim 10;

wherein the translator that creates a translation result sentence by translating a predetermined to-be-translated sentence by use of the updated translation model, the updated translation model being generated from the one or more second sentences that the device has determined to use as similar sentences of the first sentence and from a translated sentence resulting from translation of the first sentence, from which the one or more second sentences have been created, in a second language;

an evaluator that evaluates the translation result sentence created by the translator; and

a creator that creates feedback information including language information related to at least one of a language of the to-be-translated sentence and a language of the translation result sentence as well as evaluation information for the language information, according to an evaluation result from the evaluator.