PENGGUNAAN SPEECH-TO-TEXT DAN MODEL NLP UNTUK PENYEDIAAN RANGKUMAN OTOMATIS SEBAGAI DUKUNGAN PEMBELAJARAN DARING

    Syifa Nur Pratiwi, - and Asyifa Imanda Septiana, - and Indira Syawanodya, - (2025) PENGGUNAAN SPEECH-TO-TEXT DAN MODEL NLP UNTUK PENYEDIAAN RANGKUMAN OTOMATIS SEBAGAI DUKUNGAN PEMBELAJARAN DARING. S1 thesis, Universitas Pendidikan Indonesia.

    Abstract

    Pembelajaran daring pasca pandemi COVID-19 tetap dipertahankan dalam bentuk kuliah hybrid, seminar, maupun pelatihan profesional. Namun, rekaman sesi daring yang tersedia sering bersifat pasif dan memerlukan peninjauan manual, sehingga menimbulkan tantangan dalam dokumentasi dan retensi informasi. Penelitian ini mengusulkan prototipe sistem rangkuman otomatis berbasis Automatic Speech Recognition (ASR) dan Natural Language Processing (NLP). Sistem menggunakan Whisper Medium sebagai ASR untuk mentranskripsi audio pembelajaran daring ke teks, serta T5 Bahasa Indonesia untuk merangkum hasil transkripsi. Evaluasi dilakukan dengan Word Error Rate (WER) dan ROUGE. Hasil menunjukkan Whisper Medium memberikan akurasi transkripsi cukup baik (WER 28,51%), sementara model T5 menghasilkan ringkasan relevan dengan skor ROUGE-1 sebesar 67,46%. Pada pengujian durasi panjang (4 jam), WER meningkat hingga 89,83% akibat noise dan pemotongan rekaman, menandakan sistem lebih optimal untuk segmen menengah (10–20 menit). Temuan ini menunjukkan potensi sistem sebagai alat bantu pembelajaran daring, meskipun perlu diwaspadai bahwa penggunaan rangkuman otomatis secara berlebihan dapat menurunkan keterlibatan aktif mahasiswa dalam memahami materi. ----------- Online learning, although offering flexibility, faces significant challenges in terms of documentation and information retention. The availability of recorded sessions is often insufficient, as they are passive and require time-consuming manual review. This issue remains relevant even in the post-COVID era, where online learning is still used in hybrid models and distance education. Moreover, the adoption of automated tools raises a dilemma: while they can improve efficiency, there are concerns that students may become overly reliant on them. To address these challenges, this research proposes the development of an automatic summarization prototype that integrates Automatic Speech Recognition (ASR) and Natural Language Processing (NLP). The system employs Whisper as the ASR model to transcribe Indonesian audio into text, and T5 Bahasa Indonesia as the NLP model to generate summaries. The evaluation results show that Whisper Medium achieved a Word Error Rate (WER) of 28.51%, which falls into the “moderately accurate” category (10–30%). In contrast, Whisper Tiny produced a much higher WER of 74.40%, indicating poor transcription quality. For summarization, the T5 model reached ROUGE-1 of 67.46% and ROUGE-L of 55.90%, demonstrating good coverage and coherence. However, in longer recordings (up to 4 hours), the WER increased significantly to 89.83% due to noise and segmented recording, reducing transcription accuracy. These findings suggest that while the system is effective in generating concise summaries for shorter sessions, improvements in handling noise and long-duration data are required. This research highlights the potential of ASR and NLP integration as a supportive tool for online learning in Indonesia, particularly in enhancing information retention and accessibility.

    [thumbnail of S_RPL_2109890_Title.pdf] Text
    S_RPL_2109890_Title.pdf

    Download (852kB)
    [thumbnail of S_RPL_2109890_Chapter1.pdf] Text
    S_RPL_2109890_Chapter1.pdf

    Download (321kB)
    [thumbnail of S_RPL_2109890_Chapter2.pdf] Text
    S_RPL_2109890_Chapter2.pdf
    Restricted to Staf Perpustakaan

    Download (603kB)
    [thumbnail of S_RPL_2109890_Chapter3.pdf] Text
    S_RPL_2109890_Chapter3.pdf

    Download (617kB)
    [thumbnail of S_RPL_2109890_Chapter4.pdf] Text
    S_RPL_2109890_Chapter4.pdf
    Restricted to Staf Perpustakaan

    Download (464kB)
    [thumbnail of S_RPL_2109890_Chapter5.pdf] Text
    S_RPL_2109890_Chapter5.pdf

    Download (284kB)
    [thumbnail of S_RPL_2109890_Appendix.pdf] Text
    S_RPL_2109890_Appendix.pdf
    Restricted to Staf Perpustakaan

    Download (11MB)
    Official URL: https://repository.upi.edu/
    Item Type: Thesis (S1)
    Additional Information: https://scholar.google.com/citations?view_op=new_profile&hl=en ID SINTA Dosen Pembimbing: Asyifa Imanda Septiana : 6681802 Indira Syawanodya : 6681751
    Uncontrolled Keywords: Pembelajaran Daring, Automatic Speech Recognition, Natural Language Processing, Whisper, T5 Bahasa Indonesia, Online Learning, Automatic Speech Recognition, Natural Language Processing, Whisper, T5 for Indonesia.
    Subjects: L Education > L Education (General)
    Q Science > QA Mathematics > QA75 Electronic computers. Computer science
    Q Science > QA Mathematics > QA76 Computer software
    Divisions: UPI Kampus cibiru > S1 Rekayasa Perangkaat Lunak
    Depositing User: Syifa Nur Pratiwi
    Date Deposited: 18 Sep 2025 08:23
    Last Modified: 18 Sep 2025 08:23
    URI: http://repository.upi.edu/id/eprint/137446

    Actions (login required)

    View Item View Item