Indonesian Friday Sermon Transcript Dataset from YouTube and Web Sources

This dataset presents a comprehensive collection of 213 Indonesian Friday sermon (khutbah Jumat) transcripts sourced from popular YouTube Islamic channels during 2025-2026, totaling over 2.6 million characters of authentic religious discourse. Each entry includes complete verbatim transcripts alongside rich metadata: sermon titles, YouTube URLs, viewer counts (ranging from 186 to 484K views), and video durations (averaging 18:43 minutes). The Excel file (1.07 MB) captures diverse themes including eschatology (death and afterlife), practical Islamic ethics (patience, charity, family relations), spiritual development (taqwa, repentance), and contemporary issues (Ramadan preparation, social media). Collected via automated speech-to-text with manual verification, this corpus is ideal for NLP research—particularly text summarization (mBART-50/IndoBERT fine-tuning), topic modeling, sentiment analysis of religious speech, and speech-to-text validation in Indonesian dialects. Licensed under CC BY 4.0, it serves as a valuable baseline for computational linguistics and Islamic studies, enabling ROUGE metric benchmarking and transformer model pre-training on formal religious oratory.