Transcription of Informatics Final Project Seminar Recordings via Speech-to-Text

Abstract

Bandung State Polytechnic has implemented project-oriented problem-based learning as its new educational approach. Project 6, or Final Project, is a course that communicates student learning results. Documenting seminar events is crucial as it provides valuable resources for students to analyze seminar outcomes and address any inquiries from teachers. However, not all students are inherently proactive in documenting or recording these activities. The task of transcribing learning outcomes becomes distinct when the emphasis is placed on students. This study aims to develop and evaluate a speech-to-text model utilizing DeepSpeech for transcribing seminar presentations related to final year projects, tackling the difficulties presented by spontaneous speech patterns and specialized technical terminology in software engineering. The model is trained and assessed utilizing Word Error Rate (WER) and Character Error Rate (CER) measures. The results of this study are the development of speech-to-text systems for educational purposes, especially within project-based student-centered learning. These resulting transcriptions could benefit both students and educators by offering a searchable and analyzable account of seminar presentations and improving feedback.

Keywords

• speech to text
• final project
• deepspeech
• seminar recording