The case where the learning termination determining section 15 determines that the difference between the weight W′_N×V before the update and the weight W′_N×V after the update is smaller than the threshold as the requirement for the termination of the learning is described above, but the embodiment is not limited to this. The learning termination determining section 15 may terminate the learning when the predetermined process (iteration) is executed a predetermined number of times.

[Flowchart of Learning Process]

FIG. 6 is a diagram illustrating an example of a flowchart of a learning process according to the embodiment. The example assumes that a word space of a reference language with language resources of an amount larger than the defined amount and a word space of a target language with language resources of an amount smaller than the defined amount exist.

As illustrated in FIG. 6, the learning unit 10 determines whether or not the learning unit 10 has received a learning process request (in step S10). When the learning unit 10 determines that the learning unit 10 has not received the learning process request (No in step S10), the learning unit 10 repeats the determination process until the learning unit 10 receives the learning process request.

When the learning unit 10 determines that the learning unit 10 has received the learning process request (Yes in step S10), the learning unit 10 learns distributed representations of words in the reference language (in step S20). For example, the learning unit 10 uses the Skip-gram model of Word2Vec to learn the distributed representations of the words in the reference language.