datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
sermons
Contemporary Preaching: Manuscript Sermons
324 sermons (~665,000 words) written and preached by Wilson Pruitt, a United Methodist pastor, from April 2020 to August 2026 at two Texas congregations (Bee Creek UMC, 166; Covenant UMC Austin, 158). One JSON record per sermon.
Why this dataset exists
Most public "sermon" corpora are automatic speech transcripts of YouTube video: spoken-word artifacts with recognition errors and no structure. This is the opposite: the… See the full description on the dataset page: https://huggingface.co/datasets/wrootpress/sermons.charles_spurgeon_sermons
Charles Spurgeon Public Domain Dataset
Description
This dataset contains cleaned, public-domain writings from Charles Spurgeon. It is intended for language modeling experiments.
Contents
Author: Charles Spurgeon
Works included: Spurgeon's Sermons Volumes 01-63
Format: plain text, UTF-8
Size
Total tokens (corpus): 24,073,429
Train (90%): 21,658,949 tokens
Validation (5%): 1,207,406 tokens
Test (5%): 1,207,074 tokens
Number of entries (files): 3,492… See the full description on the dataset page: https://huggingface.co/datasets/BobTheUnicorn987/charles_spurgeon_sermons.ekklesiaoradea_sermonsunitarian-universalist-sermons
6900 transcripts
44 churches
timeframe: 2010-2022
Denomination: Unitarian Universalist, USA
Dataset structure
church (church name or website)
source (mp3 file)
text
sentences (count)
errors (number of sentences skipped because could not understand audio, or just long pauses skipped)
duration (in seconds)
Dataset creation
see notebook in files
whisper_tp_sermons