Pengembangan Model Pembelajaran Mesin Alih Bentuk Suara ke Teks Menggunakan Algoritma Support Vector Machine dengan Library SVM dari Scikit-LearnTedi Lesmana Marselino

Abstract

The purpose of this research is to develop a speech-to-text Machine Learning model. It uses audio files in Indonesian Language as the model’s input. The method used in this research was Incremental with one Increment. The development process uses SVM Library from Scikit-Learn which implementing Support Vector Machine (SVM) algorithm. And the validation process uses cross-validation function to measure model validation score. This research uses Python as a programming language. The final result from this research is a model with low validation score, which is 2.5%. This happened because of algorithm error which causing the audio’s array data being overwritten so only 1 last index from each audio file was processed by the algorithm. In the model implementation process, the model can predict only 1 from 50 words correctly so the accuracy of the model through real data was 2%.