Translation apparatus

• US 10,055,404 B2
• Filed: 12/27/2016
• Issued: 08/21/2018
• Est. Priority Date: 01/29/2016
• 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
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Le, Thuykhanh
• Application Number:
  US15/391,354
• Publication Number:
  US 20170220562A1
• Time in Patent Office: 602 Days
• Field of Search: None
• CPC Class Codes: G06F 40/289 Phrasal analysis, e.g. finite state techniques or chunking
• CPC Class Codes: G06F 40/45 Example-based machine translation; Alignment
• CPC Class Codes: G06F 40/51 Translation evaluation
• CPC Class Codes: G06F 40/53 Processing of non-Latin text kana-to-kanji conversion G06F40/129; vowelisation G06F40/232

Abstract

Provided is a method of managing a server having stored thereon a phrase table containing phrase pairs each being a pair of a phrase in a first language and a phrase in a second language different from the first language. In the method, a processor receives combinations of the phrase pairs used to generate respective candidate target texts which are generated by translation of a source text in the first language into a target text in the second language as candidates for the target text in the second language, and records combination information (for example, a data structure and paths) in the server, the combination information identifying, in the phrase pairs contained in the phrase table, the combinations of the phrase pairs used to generate the respective candidate target texts.

Claims

1. A method for performing machine translation using a translation apparatus having a processor, the method comprising:

receiving, using the processor, a plurality of respective candidate target texts which are generated by translation of a source text in a first language into a target text in a second language different from the first language as candidates for the target text in the second language;

specifying, using the processor, combinations of phrase pairs used to generate the respective candidate target texts among phrase pairs contained in a phrase table of a translation model, the phrase table stored in a server and containing the phrase pairs each being a pair of a phrase in the first language and a phrase in the second language; and

recording, using the processor, combination information in the server, the combination information identifying the combinations of phrase pairs,

wherein the phrase pairs contained in the phrase table of the translation model are respectively associated with scores each of which is information on probability of occurrence of the phrase pair associated therewith,

the method further comprising;

prompting a user to select a most suitable candidate target text from the candidate target texts;

using the recorded combination information, identifying the combination of the phrase pairs used to generate the candidate target text not selected in the selecting;

evaluating each of the phrase pairs forming the identified combination so as to adjust a score of the phrase pair;

updating the translation model by learning the adjusted score of the phrase pair; and

retranslating the source text using the updated translation model.

2. The method according to claim 1,

wherein the recording includes recording, as the combination information, a data structure representing the combinations of the phrase pairs used to generate the respective candidate target texts, and paths along which the respective combinations of the phrase pairs used to generate the respective candidate target texts are obtainable in the data structure.

3. The method according to claim 1,

wherein the evaluating includes comparing each of the phrase pairs forming the identified combination with elements of the selected candidate target text, and determining, based on predetermined criteria, to increase the score of the phrase pair whose score is to be increased, and to decrease the score of the phrase pair whose score is to be decreased.

4. The method according to claim 3, further comprising:

performing processing which is at least one of a first process of increasing the score of the phrase pair whose score is determined to be increased and a second process of decreasing the score of the phrase pair whose score is determined to be decreased.

5. The method according to claim 4, further comprising:

when the first process is performed in the processing, setting a first predetermined value for use in the first process according to variability between the score of the phrase pair to be subjected to the first process and the score of another phrase pair containing a same phrase in the first language as the phrase pair to be subjected to the first process,

wherein the performing the processing includes performing the first process using the first predetermined value thus set.

6. The method according to claim 4, further comprising:

when the second process is performed in the processing, setting a second predetermined value for use in the second process according to variability between the score of the phrase pair to be subjected to the second process and the score of another phrase pair containing a same phrase in the first language as the phrase pair to be subjected to the second process,

wherein the performing the processing includes performing the second process using the second predetermined value thus set.

7. The method according to claim 4, wherein

the recording includes recording translation-result-related numerical information on the candidate target texts, the translation-result-related numerical information being calculated based on the scores, the method further comprising:

retranslating the source text using the phrase table in the updated translation model, to generate again a plurality of candidate target texts as candidates for the target text;

calculating and generating the numerical information on each of the candidate target texts thus generated again, based on the scores in the phrase table updated by the processing;

determining whether a predefined criterion is satisfied or not based on a comparison between the numerical information recorded and the numerical information generated; and

performing the processing again when it is determined that the predefined criterion is satisfied.

8. The method according to claim 7, further comprising:

translating a target text from the first language to the second language using the phrase table; and

displaying the translated target text on a display.

9. A translation apparatus comprising:

a recording medium having a phrase table stored thereon, the phrase table containing phrase pairs each being a pair of a phrase in a first language and a phrase in a second language different from the first language;

a processor; and

a memory having a computer program stored thereon, the computer program causing the processor to execute machine translation operations including;

receiving a plurality of respective candidate target texts which are generated by translation of a source text in the first language into a target text in the second language as candidates for the target text in the second language;

specifying combinations of phrase pairs used to generate the respective candidate target texts among phrase pairs contained in a phrase table of a translation model, the phrase table stored in a server and containing the phrase pairs each being a pair of a phrase in the first language and a phrase in the second language; and

recording combination information in the recording medium, the combination information identifying the combinations of phrase pairs,

wherein the phrase pairs contained in the phrase table of the translation model are respectively associated with scores each of which is information on probability of occurrence of the phrase pair associated therewith, and

wherein the machine translation operations further include;

prompting a user to select a most suitable candidate target text from the candidate target texts;

using the recorded combination information, identifying the combination of the phrase pairs used to generate the candidate target text not selected in the selecting; and

evaluating each of the phrase pairs forming the identified combination so as to adjust a score of the phrase pair;

updating the translation model by learning the adjusted score of the phrase pair; and

retranslating the source text using the updated translation model.

10. A translation apparatus comprising:

a processor;

a recording medium having a phrase table stored thereon, the phrase table containing phrase pairs each being a pair of a phrase in a first language and a phrase in a second language different from the first language; and

a medium having a computer program stored thereon, the computer program causing the processor to execute machine translation operations including;

generating a plurality of candidate target texts as candidates for a target text using the phrase table to translate a source text in the first language into the target text in the second language;

specifying combinations of phrase pairs used to generate the respective candidate target texts among phrase pairs contained in a phrase table of a translation model, the phrase table stored in a server and containing the phrase pairs each being a pair of a phrase in the first language and a phrase in the second language; and

recording combination information in the recording medium, the combination information identifying the combinations of phrase pairs,

wherein the phrase pairs contained in the phrase table of the translation model are respectively associated with scores each of which is information on probability of occurrence of the phrase pair associated therewith, and

wherein the machine translation operations further include;

prompting a user to select a most suitable candidate target text from the candidate target texts;

using the recorded combination information, identifying the combination of the phrase pairs used to generate the candidate target text not selected in the selecting;

evaluating each of the phrase pairs forming the identified combination so as to adjust a score of the phrase pair;

updating the translation model by learning the adjusted score of the phrase pair; and

retranslating the source text using the updated translation model.

11. A non-transitory recording medium having a computer program stored thereon, the computer program causing a processor of a translation apparatus to execute machine translation operations comprising:

receiving a plurality of respective candidate target texts which are generated by translation of a source text in a first language into a target text in a second language different from the first language as candidates for the target text in the second language;

specifying combinations of phrase pairs used to generate the respective candidate target texts among phrase pairs contained in a phrase table of a translation model, the phrase table stored in a server and containing the phrase pairs each being a pair of a phrase in the first language and a phrase in the second language; and

recording combination information in the server, the combination information which identifies the combinations of phrase pairs,

wherein the phrase pairs contained in the phrase table of the translation model are respectively associated with scores each of which is information on probability of occurrence of the phrase pair associated therewith, and

wherein the machine translation operations further include;

prompting a user to select a most suitable candidate target text from the candidate target texts;

using the recorded combination information, identifying the combination of the phrase pairs used to generate the candidate target text not selected in the selecting;

evaluating each of the phrase pairs forming the identified combination so as to adjust a score of the phrase pair;

updating the translation model by learning the adjusted score of the phrase pair; and

retranslating the source text using the updated translation model.