Multilingual Communication via Best-Balanced Machine Translation

It is known that creative ideas are often generated by correspondents from different cultures, but it remains true that distance still matters due to the language barrier. To enhance multilingual communication, this paper proposes the model of best-balanced machine translation. Our model is based on the quality of messages among participants (assumed to have different levels of language skill) and takes not only machine translation quality but also users’ language skill into account. We provide a method to select languages to be used with machine translators, and a way of creating the best-balanced communication environment. Many studies have addressed machine translation with the goal of helping the non-native speaker to understand what was said. Our approach is totally differently, since we focus on helping the non-native speaker by enhancing the opportunity to join in the conversation. We conduct an experiment and find that this model allows machine translation technologies to benefit multilingual communication while making best use of the participants’ different language skills. The proposed model addresses the talkativeness of the participants. It also improves communication by reducing serious machine translation errors and the number of conversation breakdowns.