Multilingual Speech Translator for Medical Consultation

The speaking at cross purposes is one of the common problems in elderly care scenarios, especially the cross-language conversation in medical consultations. Constructing a multilingual speech translation service for medical consultation can improve the efficiency of cross-language conversations. This study focuses on proposed a multilingual speech translator for Chinese, Indonesian and Taiwanese processing. In this paper, an under-resourced Taiwanese ASR model was constructed by the collected Tailuo pinyin dataset. The experimental results show the proposed Taiwanese ASR model achieved good performance on Taiwanese speech recognition. Furthermore, the Taiwanese ASR model and Indonesian ASR model are integrated into a speech translation system to construct a speech-to-speech translator for medical consultation tasks. The proposed multilingual speech translator demonstrated the practicality and functionality for auxiliary application tasks such as cross-language conversation.