Thai Horror Podcast Narrative Skeleton Dataset

Derived narrative-schema dataset for 200 Thai horror stories drawn fromthree YouTube podcast channels (เล่าไปหลอนไป, หลอนสเปเชียล, 108เรื่องหลอน). Each story is represented as a seven-field narrativeskeleton (setting, trigger, escalation pattern, entity type, climaxaction, resolution type, sacred marker), extracted via a large languagemodel pipeline and validated through a three-layer schema-conformancecheck (198/200 stories, 99.0% clean). Also includes FP-Growth frequentitemset results, K-Modes cluster labels (K=5), and YouTube engagementstatistics (view/like/comment percentiles) used for the popularityanalysis. Raw audio, video, and full transcripts are NOT included, as these remainthird-party copyrighted content owned by the respective YouTube channels.YouTube video IDs are preserved so records can be relinked to theirpublic source. Accompanies the paper "Finite Skeletons, Ambiguous Endings: ComputationalEvidence for Recurring Narrative Templates in Contemporary Thai HorrorPodcasts" (submitted to Digital Scholarship in the Humanities, 2026).