I report a situational prompt translation practice system for language learning. The system first builds a database holding many of short sentences. After that, it selects a sentence from the sentences through a quasi-random process and then displays it on the screen. The learner promptly and verbally translates it into the foreign language. The system employs such a method that the learner verifies how precisely the sentence has been translated by comparing the learner's translation with the display answer and the voice answer indicated by the system. The system aims at the learner who finds difficulty in promptly speaking what he/she wants to say in the foreign language. Because there is not necessarily one correct answer, the learner must have the ability to judge how precise the his/her answer is by visually and verbally checking the answer indicated by the system, which would be one of the correct answers. The adjective "situational" derives from the fact that the system has adopted such an algorithm that the occurrence probability of a short sentence which is suitable under a certain situation will be increased and thus the occurrence probability of an unsuitable sentence will be decreased, by calculating the compatibility between the property of the short sentence and the property of the user, even though the short sentence displayed on the screen has basically been selected through a quasi-random process. In the calculation above, fuzzy theory is implemented when comparing two properties mentioned above.