Speech to Text of Patient Complaints for Bahasa Indonesia

摘要

Speech to text is a challenging task in the natural language processing research area. In this paper, we proposed a preliminary study in building speech to text application of patient complaints for Bahasa Indonesia during anamnesis. This study aims to apply speech to text using TensorFlow for Bahasa Indonesia in the medical domain. The voice data used in this research were not only the words in Bahasa Indonesia but also some complaints in Javanese language. We have collected 500 voices recorded data from 10 persons. We have tested 50 data and obtained the accuracy of 64°/0. In addition, we have also performed the similarity calculation to compare between the spoken word and the generated text. From this similarity calculation, we classify our generation text data into three categories, good, fair, and bad. For future work, it is important to add more voice data, both in Bahasa Indonesia and Javanese Language.