Then, the information processing device 1 inputs, for the group B, the adjusted predetermined parameter to the single algorithm of the program, including Word2Vec, for learning a distributed representation of a word and learns a distributed representation of the word, included in the group B, in the target language. For example, the information processing device 1 learns a relative distributed representation of the uncommon word in the target language by crosschecking the uncommon word in the target language to the positions of the words included in the group A. For example, the distributed representation of the uncommon target language word included in the group B and corresponding to the word “sword” is relatively learned by crosschecking the uncommon target language word to the positions of the words included in the group A.

Thus, even when the amount of the language resources included in the word space of the target language is not sufficient, the information processing device 1 may learn distributed representations with high accuracy. In the learning of a distributed representation of a word, the case where the amount of language resources is not sufficient may not be assumed, and when the amount of language resources is not sufficient, distributed representations of words, which are language resources, may not be learned. However, even when the amount of language resources is not sufficient, the information processing device 1 may learn distributed representations with high accuracy.