Hybrid machine translation for Javanese speech levels

Javanese is a local language with the biggest number of speakers in Indonesia. However, the fact that it is characterized by a complex system of politeness means that it is perceived negatively by Javanese teenagers, who have difficulty understanding it. The hybrid corpus-based machine translation described here is designed to address this by offering a means of translating speech levels appropriately. The system embeds statistical features into a memory-based machine translation to obtain the best performance of Javanese speech levels´ translation. The evaluation shows satisfactory results; 0.83 and 90.4 for the average accuracy and quality of the translation, respectively.