Apparatus, method, and computer program product for creating data for learning word translation

A data creating method performed in a data creating apparatus that creates learning data for learning proper usage of a plurality of translation candidates corresponding to one word in a source language, the translation candidates being described in a target language, the method comprising:

• inputting a sentence in the source language;
• translating one word in the sentence into a target language based on a predetermined translation rule, and generating a plurality of translation candidates in the target language with respect to the word and likelihood expressing probability of the translation candidate;
• creating, using a processor, a plurality of pieces of correspondence information, each correspondence information including the word and one of the translation candidates;
• determining whether the likelihood of the translation candidate is larger than a predetermined threshold value;
• and
• creating the learning data in which another word in the sentence is added to the correspondence information of the translation candidate having the likelihood larger than the threshold value, wherein
• the translating includes:
• performing morphological analysis with respect to the sentence to divide the sentence into words;
• obtaining the translation candidate with respect to the divided word from a dictionary storage unit that stores a bilingual dictionary in which a word in the source language and the translation candidate with respect to the word in the source language are associated with each other;
• analyzing a dependency relationship between the divided words:
• collating the divided word and a word having the dependency relationship with the divided word, with the word in the source language and the word having the dependency relationship with the word in the source language, respectively, in the translation rule stored in a translation rule storage unit for each divided word, the translation rule storage unit storing the translation rule in which the word in the source language, the word having the dependency relationship with the word in the source language, and the translation candidate are associated with each other;
• and
• obtaining a matched translation candidate associated with the word in the source language in the translation rule from the translation rule storage unit.

A non-transitory computer readable medium including programmed instructions for creating learning data for learning proper usage of a plurality of translation candidates corresponding to one word in a source language, the translation candidates being described in a target language, wherein the instructions, when executed by a computer, cause the computer to perform:

• inputting a sentence in the source language;
• translating one word in the sentence into a target language based on a predetermined translation rule, and generating a plurality of translation candidates in the target language with respect to the word and likelihood expressing probability of the translation candidate;
• creating a plurality of pieces of correspondence information, each correspondence information including the word and one of the translation candidates;
• determining whether the likelihood of the translation candidate is larger than a predetermined threshold value;
• and
• creating the learning data in which another word in the sentence is added to the correspondence information of the translation candidate having the likelihood larger than the threshold value, wherein
• the translating includes:
• performing morphological analysis with respect to the sentence to divide the sentence into words;
• obtaining the translation candidate with respect to the divided word from a dictionary storage unit that stores a bilingual dictionary in which a word in the source language and the translation candidate with respect to the word in the source language are associated with each other;
• analyzing a dependency relationship between the divided words;
• collating the divided word and a word having the dependency relationship with the divided word, with the word in the source language and the word having the dependency relationship with the word in the source language, respectively, in the translation rule stored in a translation rule storage unit for each divided word, the translation rule storage unit storing the translation rule in which the word in the source language, the word having the dependency relationship with the word in the source language, and the translation candidate are associated with each other;
• and
• obtaining a matched translation candidate associated with the word in the source language in the translation rule from the translation rule storage unit.