PURPOSE: To obtain the subject language learning machine capable of producing and displaying a word by comparing a voice input from the microphone, word pronouncing data which is previously stored, and word translation display data with each other.

CONSTITUTION: While the write mode switch W is "on", the voice input is made a formant pattern, normalized, and is transferred to the formant pattern normalized data unit LM to store the human voice. While the switch W is "off", a voice input, if given, is processed in the same way, that is, it is normalized to a formant pattern and is compared with the data of the normalized data unit LM, the voice input is recognized. The retrieval unit JM retrieves the data area of both the word translation memory DM and the word pronunciation composer VM according to the reregnized voice input, a word corresponding to the voice input is displayed, and is pronounced by means of the loudspeaker SP.