Adding an Emotions Filter to Javanese Text-to-Speech System
Edy Mulyanto, Eko Mulyanto Yuniarno, Mauridhi Hery Purnomo · 2018
One way to interact, humans use speech. Text-to-speech system (TTS) is the process of converting text into speech. Some TTS applications in the community are visual chatbot applications, screen readers, and digital talking books for the blind. The object of this research is the Javanese language, the addition of an emotional filter to the Javanese language TTS with an automatic syllabification and phonetic transcription system. The addition of an emotional filter uses the prosody manipulation method and predetermined rate factor. Perception test to test the emotional filter, while Syllable Error Rate (SER) to test the accuracy of the syllabification system and phonetic transcription. The Mean Opinion Score (MOS) is used to evaluate the level of naturalness of speech, while the Word Error Rate (WER) is to measure the performance of speech clarity. SER test shows a value of 0.985%, the WER test produces a value of 25.03% and a MOS score of 3.60 obtained from 15 respondents.